Sumit Kumar
Published © GPL3+

Kheti Kavacha - A tiny Machine Vision powered AI Scarecrow

The next-gen device to detect intrusion of animals into farms, and raises an alarm — a combination of light and sound — to scare them away.

IntermediateFull instructions providedOver 1 day105
Kheti Kavacha - A tiny Machine Vision powered AI Scarecrow

Things used in this project

Hardware components

Luxonis LUX-ESP32
×1
Raspberry Pi 3 Model B
Raspberry Pi 3 Model B
×1
Camera Module V2
Raspberry Pi Camera Module V2
×1
MATRIX Creator
MATRIX Labs MATRIX Creator
×1
SparkFun Lithium Ion Battery Pack - 10Ah (3A/1A USB Ports)
×1

Software apps and online services

Edge Impulse Studio
Edge Impulse Studio
Raspbian
Raspberry Pi Raspbian

Story

Read more

Custom parts and enclosures

Stage 2

Enclosure

Schematics

Model Training Labeled Images and XML files

POC stage 2

Dashboard Mockup

Code

POC 2 test code

Python
#!/usr/bin/env python

import cv2
import os
import sys, getopt
import signal
import time
import vlc
from matrix_lite import led
from matrix_lite import gpio
from edge_impulse_linux.image import ImageImpulseRunner

runner = None
show_camera = True

def now():
    return round(time.time() * 1000)

def get_webcams():
    port_ids = []
    for port in range(5):
        print("Looking for a camera in port %s:" %port)
        camera = cv2.VideoCapture(port)
        if camera.isOpened():
            ret = camera.read()[0]
            if ret:
                backendName =camera.getBackendName()
                w = camera.get(3)
                h = camera.get(4)
                print("Camera %s (%s x %s) found in port %s " %(backendName,h,w, port))
                port_ids.append(port)
            camera.release()
    return port_ids

def sigint_handler(sig, frame):
    print('Interrupted')
    if (runner):
        runner.stop()
    sys.exit(0)

signal.signal(signal.SIGINT, sigint_handler)

def help():
    print('python classify.py <path_to_model.eim> <Camera port ID, only required when more than 1 camera is present>')

def strobe_lights(label):
  # white led colour flashing (R,G,B,W)
  if label == 'elephant':
    led.set((0,0,0,0))
    time.sleep(5)
    led.set('black')
  else:
    led.set('black')
  
def sound_playback(label):
  if label == 'elephant':
    p = vlc.MediaPlayer(audio_file_path)  #provide some audio files
    p.play()
  else:
    p.stop()

def main(argv):
    try:
        opts, args = getopt.getopt(argv, "h", ["--help"])
    except getopt.GetoptError:
        help()
        sys.exit(2)

    for opt, arg in opts:
        if opt in ('-h', '--help'):
            help()
            sys.exit()

    if len(args) == 0:
        help()
        sys.exit(2)

    model = args[0]

    dir_path = os.path.dirname(os.path.realpath(__file__))
    modelfile = os.path.join(dir_path, model)

    print('MODEL: ' + modelfile)

    with ImageImpulseRunner(modelfile) as runner:
        try:
            model_info = runner.init()
            print('Loaded runner for "' + model_info['project']['owner'] + ' / ' + model_info['project']['name'] + '"')
            labels = model_info['model_parameters']['labels']
            if len(args)>= 2:
                videoCaptureDeviceId = int(args[1])
            else:
                port_ids = get_webcams()
                if len(port_ids) == 0:
                    raise Exception('Cannot find any webcams')
                if len(args)<= 1 and len(port_ids)> 1:
                    raise Exception("Multiple cameras found. Add the camera port ID as a second argument to use to this script")
                videoCaptureDeviceId = int(port_ids[0])

            camera = cv2.VideoCapture(videoCaptureDeviceId)
            ret = camera.read()[0]
            if ret:
                backendName = camera.getBackendName()
                w = camera.get(3)
                h = camera.get(4)
                print("Camera %s (%s x %s) in port %s selected." %(backendName,h,w, videoCaptureDeviceId))
                camera.release()
            else:
                raise Exception("Couldn't initialize selected camera.")

            next_frame = 0 # limit to ~10 fps here

            for res, img in runner.classifier(videoCaptureDeviceId):
                if (next_frame > now()):
                    time.sleep((next_frame - now()) / 1000)

                # print('classification runner response', res)

                if "classification" in res["result"].keys():
                    print('Result (%d ms.) ' % (res['timing']['dsp'] + res['timing']['classification']), end='')
                    for label in labels:
                        score = res['result']['classification'][label]
                        print('%s: %.2f\t' % (label, score), end='')
                        if label == 'elephant':
                          strobe_lights(label)
                          sound_playback(label)
                    print('', flush=True)

                elif "bounding_boxes" in res["result"].keys():
                    print('Found %d bounding boxes (%d ms.)' % (len(res["result"]["bounding_boxes"]), res['timing']['dsp'] + res['timing']['classification']))
                    for bb in res["result"]["bounding_boxes"]:
                        print('\t%s (%.2f): x=%d y=%d w=%d h=%d' % (bb['label'], bb['value'], bb['x'], bb['y'], bb['width'], bb['height']))
                        img = cv2.rectangle(img, (bb['x'], bb['y']), (bb['x'] + bb['width'], bb['y'] + bb['height']), (255, 0, 0), 1)

                if (show_camera):
                    cv2.imshow('edgeimpulse', cv2.cvtColor(img, cv2.COLOR_RGB2BGR))
                    if cv2.waitKey(1) == ord('q'):
                        break

                next_frame = now() + 100
        finally:
            if (runner):
                runner.stop()

if __name__ == "__main__":
   main(sys.argv[1:])

HTTPS file server example

C/C++
/* HTTP File Server Example
   This example code is in the Public Domain (or CC0 licensed, at your option.)
   Unless required by applicable law or agreed to in writing, this
   software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
   CONDITIONS OF ANY KIND, either express or implied.
*/

#include <main.hpp>

#include <stdio.h>
#include <string.h>
#include <sys/param.h>
#include <sys/unistd.h>
#include <sys/stat.h>
#include <dirent.h>

#include "esp_err.h"
#include "esp_log.h"

#include "esp_vfs.h"
#include "esp_spiffs.h"
#include "esp_http_server.h"

/* Max length a file path can have on storage */
#define FILE_PATH_MAX (ESP_VFS_PATH_MAX + CONFIG_SPIFFS_OBJ_NAME_LEN)

/* Max size of an individual file. Make sure this
 * value is same as that set in upload_script.html */
#define MAX_FILE_SIZE   (1000*1024) // 1000 KB
#define MAX_FILE_SIZE_STR "1000KB"

/* Scratch buffer size */
#define SCRATCH_BUFSIZE  8192

struct file_server_data {
    /* Base path of file storage */
    char base_path[ESP_VFS_PATH_MAX + 1];

    /* Scratch buffer for temporary storage during file transfer */
    char scratch[SCRATCH_BUFSIZE];
};

static const char *TAG = "file_server";

/* Handler to redirect incoming GET request for /index.html to /
 * This can be overridden by uploading file with same name */
static esp_err_t index_html_get_handler(httpd_req_t *req)
{
    httpd_resp_set_status(req, "307 Temporary Redirect");
    httpd_resp_set_hdr(req, "Location", "/");
    httpd_resp_send(req, NULL, 0);  // Response body can be empty
    return ESP_OK;
}

/* Handler to respond with an icon file embedded in flash.
 * Browsers expect to GET website icon at URI /favicon.ico.
 * This can be overridden by uploading file with same name */
static esp_err_t favicon_get_handler(httpd_req_t *req)
{
    extern const unsigned char favicon_ico_start[] asm("_binary_favicon_ico_start");
    extern const unsigned char favicon_ico_end[]   asm("_binary_favicon_ico_end");
    const size_t favicon_ico_size = (favicon_ico_end - favicon_ico_start);
    httpd_resp_set_type(req, "image/x-icon");
    httpd_resp_send(req, (const char *)favicon_ico_start, favicon_ico_size);
    return ESP_OK;
}

/* Send HTTP response with a run-time generated html consisting of
 * a list of all files and folders under the requested path.
 * In case of SPIFFS this returns empty list when path is any
 * string other than '/', since SPIFFS doesn't support directories */
static esp_err_t http_resp_dir_html(httpd_req_t *req, const char *dirpath)
{
    char entrypath[FILE_PATH_MAX];
    char entrysize[16];
    const char *entrytype;

    struct dirent *entry;
    struct stat entry_stat;

    DIR *dir = opendir(dirpath);
    const size_t dirpath_len = strlen(dirpath);

    /* Retrieve the base path of file storage to construct the full path */
    strlcpy(entrypath, dirpath, sizeof(entrypath));

    if (!dir) {
        ESP_LOGE(TAG, "Failed to stat dir : %s", dirpath);
        /* Respond with 404 Not Found */
        httpd_resp_send_err(req, HTTPD_404_NOT_FOUND, "Directory does not exist");
        return ESP_FAIL;
    }

    /* Send HTML file header */
    httpd_resp_sendstr_chunk(req, "<!DOCTYPE html><html><body>");

    /* Get handle to embedded file upload script */
    extern const unsigned char upload_script_start[] asm("_binary_upload_script_html_start");
    extern const unsigned char upload_script_end[]   asm("_binary_upload_script_html_end");
    const size_t upload_script_size = (upload_script_end - upload_script_start);

    /* Add file upload form and script which on execution sends a POST request to /upload */
    httpd_resp_send_chunk(req, (const char *)upload_script_start, upload_script_size);

    /* Send file-list table definition and column labels */
    httpd_resp_sendstr_chunk(req,
        "<table class=\"fixed\" border=\"1\">"
        "<col width=\"800px\" /><col width=\"300px\" /><col width=\"300px\" /><col width=\"100px\" />"
        "<thead><tr><th>Name</th><th>Type</th><th>Size (Bytes)</th><th>Delete</th></tr></thead>"
        "<tbody>");

    /* Iterate over all files / folders and fetch their names and sizes */
    while ((entry = readdir(dir)) != NULL) {
        entrytype = (entry->d_type == DT_DIR ? "directory" : "file");

        strlcpy(entrypath + dirpath_len, entry->d_name, sizeof(entrypath) - dirpath_len);
        if (stat(entrypath, &entry_stat) == -1) {
            ESP_LOGE(TAG, "Failed to stat %s : %s", entrytype, entry->d_name);
            continue;
        }
        sprintf(entrysize, "%ld", entry_stat.st_size);
        ESP_LOGI(TAG, "Found %s : %s (%s bytes)", entrytype, entry->d_name, entrysize);

        /* Send chunk of HTML file containing table entries with file name and size */
        httpd_resp_sendstr_chunk(req, "<tr><td><a href=\"");
        httpd_resp_sendstr_chunk(req, req->uri);
        httpd_resp_sendstr_chunk(req, entry->d_name);
        if (entry->d_type == DT_DIR) {
            httpd_resp_sendstr_chunk(req, "/");
        }
        httpd_resp_sendstr_chunk(req, "\">");
        httpd_resp_sendstr_chunk(req, entry->d_name);
        httpd_resp_sendstr_chunk(req, "</a></td><td>");
        httpd_resp_sendstr_chunk(req, entrytype);
        httpd_resp_sendstr_chunk(req, "</td><td>");
        httpd_resp_sendstr_chunk(req, entrysize);
        httpd_resp_sendstr_chunk(req, "</td><td>");
        httpd_resp_sendstr_chunk(req, "<form method=\"post\" action=\"/delete");
        httpd_resp_sendstr_chunk(req, req->uri);
        httpd_resp_sendstr_chunk(req, entry->d_name);
        httpd_resp_sendstr_chunk(req, "\"><button type=\"submit\">Delete</button></form>");
        httpd_resp_sendstr_chunk(req, "</td></tr>\n");
    }
    closedir(dir);

    /* Finish the file list table */
    httpd_resp_sendstr_chunk(req, "</tbody></table>");

    /* Send remaining chunk of HTML file to complete it */
    httpd_resp_sendstr_chunk(req, "</body></html>");

    /* Send empty chunk to signal HTTP response completion */
    httpd_resp_sendstr_chunk(req, NULL);
    return ESP_OK;
}

#define IS_FILE_EXT(filename, ext) \
    (strcasecmp(&filename[strlen(filename) - sizeof(ext) + 1], ext) == 0)

/* Set HTTP response content type according to file extension */
static esp_err_t set_content_type_from_file(httpd_req_t *req, const char *filename)
{
    if (IS_FILE_EXT(filename, ".pdf")) {
        return httpd_resp_set_type(req, "application/pdf");
    } else if (IS_FILE_EXT(filename, ".html")) {
        return httpd_resp_set_type(req, "text/html");
    } else if (IS_FILE_EXT(filename, ".jpeg")) {
        return httpd_resp_set_type(req, "image/jpeg");
    } else if (IS_FILE_EXT(filename, ".ico")) {
        return httpd_resp_set_type(req, "image/x-icon");
    }
    /* This is a limited set only */
    /* For any other type always set as plain text */
    return httpd_resp_set_type(req, "text/plain");
}

/* Copies the full path into destination buffer and returns
 * pointer to path (skipping the preceding base path) */
static const char* get_path_from_uri(char *dest, const char *base_path, const char *uri, size_t destsize)
{
    const size_t base_pathlen = strlen(base_path);
    size_t pathlen = strlen(uri);

    const char *quest = strchr(uri, '?');
    if (quest) {
        pathlen = MIN(pathlen, quest - uri);
    }
    const char *hash = strchr(uri, '#');
    if (hash) {
        pathlen = MIN(pathlen, hash - uri);
    }

    if (base_pathlen + pathlen + 1 > destsize) {
        /* Full path string won't fit into destination buffer */
        return NULL;
    }

    /* Construct full path (base + path) */
    strcpy(dest, base_path);
    strlcpy(dest + base_pathlen, uri, pathlen + 1);

    /* Return pointer to path, skipping the base */
    return dest + base_pathlen;
}

/* Handler to download a file kept on the server */
static esp_err_t download_get_handler(httpd_req_t *req)
{
    char filepath[FILE_PATH_MAX];
    FILE *fd = NULL;
    struct stat file_stat;

    const char *filename = get_path_from_uri(filepath, ((struct file_server_data *)req->user_ctx)->base_path,
                                             req->uri, sizeof(filepath));
    if (!filename) {
        ESP_LOGE(TAG, "Filename is too long");
        /* Respond with 500 Internal Server Error */
        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Filename too long");
        return ESP_FAIL;
    }

    /* If name has trailing '/', respond with directory contents */
    if (filename[strlen(filename) - 1] == '/') {
        return http_resp_dir_html(req, filepath);
    }

    if (stat(filepath, &file_stat) == -1) {
        /* If file not present on SPIFFS check if URI
         * corresponds to one of the hardcoded paths */
        if (strcmp(filename, "/index.html") == 0) {
            return index_html_get_handler(req);
        } else if (strcmp(filename, "/favicon.ico") == 0) {
            return favicon_get_handler(req);
        }
        ESP_LOGE(TAG, "Failed to stat file : %s", filepath);
        /* Respond with 404 Not Found */
        httpd_resp_send_err(req, HTTPD_404_NOT_FOUND, "File does not exist");
        return ESP_FAIL;
    }

    fd = fopen(filepath, "r");
    if (!fd) {
        ESP_LOGE(TAG, "Failed to read existing file : %s", filepath);
        /* Respond with 500 Internal Server Error */
        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to read existing file");
        return ESP_FAIL;
    }

    ESP_LOGI(TAG, "Sending file : %s (%ld bytes)...", filename, file_stat.st_size);
    set_content_type_from_file(req, filename);

    /* Retrieve the pointer to scratch buffer for temporary storage */
    char *chunk = ((struct file_server_data *)req->user_ctx)->scratch;
    size_t chunksize;
    do {
        /* Read file in chunks into the scratch buffer */
        chunksize = fread(chunk, 1, SCRATCH_BUFSIZE, fd);

        if (chunksize > 0) {
            /* Send the buffer contents as HTTP response chunk */
            if (httpd_resp_send_chunk(req, chunk, chunksize) != ESP_OK) {
                fclose(fd);
                ESP_LOGE(TAG, "File sending failed!");
                /* Abort sending file */
                httpd_resp_sendstr_chunk(req, NULL);
                /* Respond with 500 Internal Server Error */
                httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to send file");
               return ESP_FAIL;
           }
        }

        /* Keep looping till the whole file is sent */
    } while (chunksize != 0);

    /* Close file after sending complete */
    fclose(fd);
    ESP_LOGI(TAG, "File sending complete");

    /* Respond with an empty chunk to signal HTTP response completion */
    httpd_resp_send_chunk(req, NULL, 0);
    return ESP_OK;
}

/* Handler to upload a file onto the server */
static esp_err_t upload_post_handler(httpd_req_t *req)
{
    char filepath[FILE_PATH_MAX];
    FILE *fd = NULL;
    struct stat file_stat;

    /* Skip leading "/upload" from URI to get filename */
    /* Note sizeof() counts NULL termination hence the -1 */
    const char *filename = get_path_from_uri(filepath, ((struct file_server_data *)req->user_ctx)->base_path,
                                             req->uri + sizeof("/upload") - 1, sizeof(filepath));
    if (!filename) {
        /* Respond with 500 Internal Server Error */
        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Filename too long");
        return ESP_FAIL;
    }

    /* Filename cannot have a trailing '/' */
    if (filename[strlen(filename) - 1] == '/') {
        ESP_LOGE(TAG, "Invalid filename : %s", filename);
        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Invalid filename");
        return ESP_FAIL;
    }

    if (stat(filepath, &file_stat) == 0) {
        ESP_LOGE(TAG, "File already exists : %s", filepath);
        /* Respond with 400 Bad Request */
        httpd_resp_send_err(req, HTTPD_400_BAD_REQUEST, "File already exists");
        return ESP_FAIL;
    }

    /* File cannot be larger than a limit */
    if (req->content_len > MAX_FILE_SIZE) {
        ESP_LOGE(TAG, "File too large : %d bytes", req->content_len);
        /* Respond with 400 Bad Request */
        httpd_resp_send_err(req, HTTPD_400_BAD_REQUEST,
                            "File size must be less than "
                            MAX_FILE_SIZE_STR "!");
        /* Return failure to close underlying connection else the
         * incoming file content will keep the socket busy */
        return ESP_FAIL;
    }

    fd = fopen(filepath, "w");
    if (!fd) {
        ESP_LOGE(TAG, "Failed to create file : %s", filepath);
        /* Respond with 500 Internal Server Error */
        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to create file");
        return ESP_FAIL;
    }

    ESP_LOGI(TAG, "Receiving file : %s...", filename);

    /* Retrieve the pointer to scratch buffer for temporary storage */
    char *buf = ((struct file_server_data *)req->user_ctx)->scratch;
    int received;

    /* Content length of the request gives
     * the size of the file being uploaded */
    int remaining = req->content_len;

    while (remaining > 0) {

        ESP_LOGI(TAG, "Remaining size : %d", remaining);
        /* Receive the file part by part into a buffer */
        if ((received = httpd_req_recv(req, buf, MIN(remaining, SCRATCH_BUFSIZE))) <= 0) {
            if (received == HTTPD_SOCK_ERR_TIMEOUT) {
                /* Retry if timeout occurred */
                continue;
            }

            /* In case of unrecoverable error,
             * close and delete the unfinished file*/
            fclose(fd);
            unlink(filepath);

            ESP_LOGE(TAG, "File reception failed!");
            /* Respond with 500 Internal Server Error */
            httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to receive file");
            return ESP_FAIL;
        }

        /* Write buffer content to file on storage */
        if (received && (received != fwrite(buf, 1, received, fd))) {
            /* Couldn't write everything to file!
             * Storage may be full? */
            fclose(fd);
            unlink(filepath);

            ESP_LOGE(TAG, "File write failed!");
            /* Respond with 500 Internal Server Error */
            httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to write file to storage");
            return ESP_FAIL;
        }

        /* Keep track of remaining size of
         * the file left to be uploaded */
        remaining -= received;
    }

    /* Close file upon upload completion */
    fclose(fd);
    ESP_LOGI(TAG, "File reception complete");

    /* Redirect onto root to see the updated file list */
    httpd_resp_set_status(req, "303 See Other");
    httpd_resp_set_hdr(req, "Location", "/");
    httpd_resp_sendstr(req, "File uploaded successfully");
    return ESP_OK;
}

/* Handler to delete a file from the server */
static esp_err_t delete_post_handler(httpd_req_t *req)
{
    char filepath[FILE_PATH_MAX];
    struct stat file_stat;

    /* Skip leading "/delete" from URI to get filename */
    /* Note sizeof() counts NULL termination hence the -1 */
    const char *filename = get_path_from_uri(filepath, ((struct file_server_data *)req->user_ctx)->base_path,
                                             req->uri  + sizeof("/delete") - 1, sizeof(filepath));
    if (!filename) {
        /* Respond with 500 Internal Server Error */
        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Filename too long");
        return ESP_FAIL;
    }

    /* Filename cannot have a trailing '/' */
    if (filename[strlen(filename) - 1] == '/') {
        ESP_LOGE(TAG, "Invalid filename : %s", filename);
        httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Invalid filename");
        return ESP_FAIL;
    }

    if (stat(filepath, &file_stat) == -1) {
        ESP_LOGE(TAG, "File does not exist : %s", filename);
        /* Respond with 400 Bad Request */
        httpd_resp_send_err(req, HTTPD_400_BAD_REQUEST, "File does not exist");
        return ESP_FAIL;
    }

    ESP_LOGI(TAG, "Deleting file : %s", filename);
    /* Delete file */
    unlink(filepath);

    /* Redirect onto root to see the updated file list */
    httpd_resp_set_status(req, "303 See Other");
    httpd_resp_set_hdr(req, "Location", "/");
    httpd_resp_sendstr(req, "File deleted successfully");
    return ESP_OK;
}


//-------------------------------------------------------------------------------------------
/* Handler to getframe from depthai */
static esp_err_t get_frame_handler(httpd_req_t *req)
{
    write_jpeg = true;
    while(write_jpeg){
        // block for jpeg write.
        usleep(100000);
    }

    /* Redirect onto root to see the updated file list */
    httpd_resp_set_status(req, "303 See Other");
    httpd_resp_set_hdr(req, "Location", "/");
    httpd_resp_sendstr(req, "File deleted successfully");
    return ESP_OK;
}

//-------------------------------------------------------------------------------------------


/* Function to start the file server */
esp_err_t start_file_server(const char *base_path)
{
    static struct file_server_data *server_data = NULL;

    /* Validate file storage base path */
    if (!base_path || strcmp(base_path, "/spiffs") != 0) {
        ESP_LOGE(TAG, "File server presently supports only '/spiffs' as base path");
        return ESP_ERR_INVALID_ARG;
    }

    if (server_data) {
        ESP_LOGE(TAG, "File server already started");
        return ESP_ERR_INVALID_STATE;
    }

    /* Allocate memory for server data */
    server_data = (file_server_data*) calloc(1, sizeof(struct file_server_data));
    if (!server_data) {
        ESP_LOGE(TAG, "Failed to allocate memory for server data");
        return ESP_ERR_NO_MEM;
    }
    strlcpy(server_data->base_path, base_path,
            sizeof(server_data->base_path));

    httpd_handle_t server = NULL;
    httpd_config_t config = HTTPD_DEFAULT_CONFIG();

    /* Use the URI wildcard matching function in order to
     * allow the same handler to respond to multiple different
     * target URIs which match the wildcard scheme */
    config.uri_match_fn = httpd_uri_match_wildcard;

    ESP_LOGI(TAG, "Starting HTTP Server");
    if (httpd_start(&server, &config) != ESP_OK) {
        ESP_LOGE(TAG, "Failed to start file server!");
        return ESP_FAIL;
    }

    /* URI handler for getting uploaded files */
    httpd_uri_t file_download = {
        .uri       = "/*",  // Match all URIs of type /path/to/file
        .method    = HTTP_GET,
        .handler   = download_get_handler,
        .user_ctx  = server_data    // Pass server data as context
    };
    httpd_register_uri_handler(server, &file_download);

    /* URI handler for uploading files to server */
    httpd_uri_t file_upload = {
        .uri       = "/upload/*",   // Match all URIs of type /upload/path/to/file
        .method    = HTTP_POST,
        .handler   = upload_post_handler,
        .user_ctx  = server_data    // Pass server data as context
    };
    httpd_register_uri_handler(server, &file_upload);

    /* URI handler for deleting files from server */
    httpd_uri_t file_delete = {
        .uri       = "/delete/*",   // Match all URIs of type /delete/path/to/file
        .method    = HTTP_POST,
        .handler   = delete_post_handler,
        .user_ctx  = server_data    // Pass server data as context
    };
    httpd_register_uri_handler(server, &file_delete);



//----------------------------------------------------------------------------------------------------
    /* URI handler for deleting files from server */
    httpd_uri_t file_getframe = {
        .uri       = "/getframe/*",
        .method    = HTTP_POST,
        .handler   = get_frame_handler,
        .user_ctx  = server_data    // Pass server data as context
    };
    httpd_register_uri_handler(server, &file_getframe);
//----------------------------------------------------------------------------------------------------

    return ESP_OK;
}

Training Code Google Colab

Python
# -*- coding: utf-8 -*-
"""Copy of Easy_Object_Detection_With_Custom_Data_Demo_Training.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/10SOHTJLK8HA35OxZBwxKFRJx_Mqy9frz

# DepthAI Tutorial: Training an Object Detection Model with Your Own Data

<img src="https://docs.luxonis.com/images/depthai_logo.png" width="500">

Welcome to DepthAI! 

In this tutorial we will go through the basic training of an object detection model with your own annotated images.

The model you will use is a pretrained Mobilenet SSD v2 from the Tensorflow Object Detection API model zoo. The framework used for training is TensorFlow 1.15.2.
Will run through the following steps:


*   Install the libraries
*   Clone the github repo and replace the repo training data with your data from google drive
    (if you do not have own data, please see this
*   Train the model on the new images
*   Run inference on a few images to see what the model can detect
*   Convert the model to OpenVINO Intermediate Representation
*   To run the model on DepthAI modules, compile the IR obtained above to a .blob file 

You can make a copy of this tutorial: File-> Save a copy in Drive

Note: the model training can be run with the repo images of three fruits if you choose to skip the part with loading your own images

# Install Libraries
# Press Shift+Enter to run cells
Some cells are commented out (all text is green), so they will not run the code inside.
To uncomment code inside a cell, select all the code then press 'Ctrl' + '/'

###After this cell executes runtime will restart to complete the install, ignore and close the message, continue running the cells below this one
"""

# %%capture
#After this cell executes runtime will restart to finish the install, ignore and close the crash message, continue running cells starting with the one below
!pip install numpy==1.17.5;

# Commented out IPython magic to ensure Python compatibility.
# %tensorflow_version 1.x
!pip install tf_slim

# For the fruit model included in the repo below we have 240 training images
# For faster training time, images should be resized to 300x300 and then annotated
# Images should contain the objects of interest at various scales, angles, lighting conditions, locations
# For acceptable results - mAP@0.5 of 0.9 the model was trained with batch size of 24
# and 5000 steps. this takes about 1h using 2 augmentations. 
# using 5 augmentations it takes about 2h 
num_steps = 500  # A step means using a single batch of data. larger batch, less steps required
#Number of evaluation steps.
num_eval_steps = 50
#Batch size 24 is a setting that generally works well. can be changed higher or lower 
MODELS_CONFIG = {
        'ssd_mobilenet_v2': {
        'model_name': 'ssd_mobilenet_v2_coco_2018_03_29',
        'pipeline_file': 'ssd_mobilenet_v2_coco.config',
        'batch_size': 24
    }
}
selected_model = 'ssd_mobilenet_v2'

# Name of the object detection model to use.
MODEL = MODELS_CONFIG[selected_model]['model_name']

# Name of the pipline file in tensorflow object detection API.
pipeline_file = MODELS_CONFIG[selected_model]['pipeline_file']

# Training batch size fits in Colab's GPU memory for selected model.
batch_size = MODELS_CONFIG[selected_model]['batch_size']

"""## Clone the `object_detection_demo_flow` repository"""

# Commented out IPython magic to ensure Python compatibility.
repo_url = 'https://github.com/GotG/object_detection_demo_flow'
import os
# %cd /content
repo_dir_path = os.path.abspath(os.path.join('.', os.path.basename(repo_url)))
!rm -rf {repo_dir_path}
!git clone {repo_url} {repo_dir_path}
# %cd {repo_dir_path}
!git pull

"""#[Optional] If you wish to use your own data, run the cells below.
(Otherwise the model can train with the data already present in the cloned repo)

## Mount your google drive
"""

#mount your google drive.
#it will be visible in the file navigator on the left of this notebook
#there should be a folder in your drive with your data
from google.colab import drive
drive.mount('/content/gdrive')

"""Own data should be split into 80% training (train folder), 20% for validation (test folder). The train and test folder should contain images and associated .xml annotations. 

Have extra images for final model testing (final_test folder). These need not be annotated.

"""

#To train on your own data:
#Remove repo data (images) for training/testing/final testing
!rm -r /content/object_detection_demo_flow/data/images/final_test/
!rm -r /content/object_detection_demo_flow/data/images/train/
!rm -r /content/object_detection_demo_flow/data/images/test/

#copy files from gdrive to colab drive. this takes a few minutes, depending on the number of files.
#Go on the file explorer on the left of this notebook and access your gdrive. find the folders
#with your train, test and final_test images.
#Right click on each and copy the path. paste it btw the first " " in the corresponding lines  

#training folder
!cp -r "/content/gdrive/My Drive/data/train/" "/content/object_detection_demo_flow/data/images/train"
#testing folder
!cp -r "/content/gdrive/My Drive/data/test/" "/content/object_detection_demo_flow/data/images/test"
#final testing folder
!cp -r "/content/gdrive/My Drive/data/final_test/" "/content/object_detection_demo_flow/data/images/final_test"

# quick check for training data files. you can also browse to the object_detection_demo_flows
# on the left and see if they were copied
!ls /content/object_detection_demo_flow/data/images/train

"""# Install Tensorflow Object Detection API

Clone TF models which contains the Object Detection API; also install the required dependencies

"""

# Commented out IPython magic to ensure Python compatibility.
# %%capture
# %cd /content
!git clone --quiet https://github.com/tensorflow/models.git
# %cd /content/models/
!git checkout 58d19c67e1d30d905dd5c6e5092348658fed80af
!apt-get update && apt-get install -y -qq protobuf-compiler python-pil python-lxml python-tk
!pip install -q Cython contextlib2 pillow lxml matplotlib
!pip install -q pycocotools
# %cd /content/models/research
!protoc object_detection/protos/*.proto --python_out=.
import os
os.environ['PYTHONPATH'] += ':/content/models/research/:/content/models/research/slim/'
!python object_detection/builders/model_builder_test.py

"""## Prepare `tfrecord` files


"""

# Commented out IPython magic to ensure Python compatibility.
# %%capture
# %cd {repo_dir_path}
# 
# # Convert train folder annotation xml files to a single csv file,
# # generate the `label_map.pbtxt` file to `data/` directory as well.
# !python xml_to_csv.py -i data/images/train -o data/annotations/train_labels.csv -l data/annotations
# 
# # Convert test folder annotation xml files to a single csv.
# !python xml_to_csv.py -i data/images/test -o data/annotations/test_labels.csv
# 
# # Generate `train.record`
# !python generate_tfrecord.py --csv_input=data/annotations/train_labels.csv --output_path=data/annotations/train.record --img_path=data/images/train --label_map data/annotations/label_map.pbtxt
# 
# # Generate `test.record`
# !python generate_tfrecord.py --csv_input=data/annotations/test_labels.csv --output_path=data/annotations/test.record --img_path=data/images/test --label_map data/annotations/label_map.pbtxt
# 
# # Set the paths
# test_record_fname = '/content/object_detection_demo_flow/data/annotations/test.record'
# train_record_fname = '/content/object_detection_demo_flow/data/annotations/train.record'
# label_map_pbtxt_fname = '/content/object_detection_demo_flow/data/annotations/label_map.pbtxt'

"""## Download the Mobilenet SSD v2 Model"""

# Commented out IPython magic to ensure Python compatibility.
# %cd /content/models/research

import os
import shutil
import glob
import urllib.request
import tarfile
MODEL_FILE = MODEL + '.tar.gz'
DOWNLOAD_BASE = 'http://download.tensorflow.org/models/object_detection/'
DEST_DIR = '/content/models/research/pretrained_model'

if not (os.path.exists(MODEL_FILE)):
    urllib.request.urlretrieve(DOWNLOAD_BASE + MODEL_FILE, MODEL_FILE)

tar = tarfile.open(MODEL_FILE)
tar.extractall()
tar.close()

os.remove(MODEL_FILE)
if (os.path.exists(DEST_DIR)):
    shutil.rmtree(DEST_DIR)
os.rename(MODEL, DEST_DIR)
!echo {DEST_DIR}
!ls -alh {DEST_DIR}

#TF pretrained model checkpoint
fine_tune_checkpoint = os.path.join(DEST_DIR, "model.ckpt")
fine_tune_checkpoint

"""## Configuring a Training Pipeline"""

import os
pipeline_fname = os.path.join('/content/models/research/object_detection/samples/configs/', pipeline_file)

assert os.path.isfile(pipeline_fname), '`{}` not exist'.format(pipeline_fname)
def get_num_classes(pbtxt_fname):
    from object_detection.utils import label_map_util
    label_map = label_map_util.load_labelmap(pbtxt_fname)
    categories = label_map_util.convert_label_map_to_categories(
        label_map, max_num_classes=90, use_display_name=True)
    category_index = label_map_util.create_category_index(categories)
    return len(category_index.keys())

from pathlib import Path
parent = Path(label_map_pbtxt_fname).parent
!ls {parent}

import re
iou_threshold = 0.50
num_classes = get_num_classes(label_map_pbtxt_fname)
with open(pipeline_fname) as f:
    s = f.read()
with open(pipeline_fname, 'w') as f:
    
    # fine_tune_checkpoint
    s = re.sub('fine_tune_checkpoint: ".*?"',
               'fine_tune_checkpoint: "{}"'.format(fine_tune_checkpoint), s)
    
    # tfrecord files train and test.
    s = re.sub(
        '(input_path: ".*?)(train.record)(.*?")', 'input_path: "{}"'.format(train_record_fname), s)
    s = re.sub(
        '(input_path: ".*?)(val.record)(.*?")', 'input_path: "{}"'.format(test_record_fname), s)

    # label_map_path
    s = re.sub(
        'label_map_path: ".*?"', 'label_map_path: "{}"'.format(label_map_pbtxt_fname), s)

    # Set training batch_size.
    s = re.sub('batch_size: [0-9]+',
               'batch_size: {}'.format(batch_size), s)

    # Set training steps, num_steps
    s = re.sub('num_steps: [0-9]+',
               'num_steps: {}'.format(num_steps), s)
    
    # Set number of classes num_classes.
    s = re.sub('num_classes: [0-9]+',
               'num_classes: {}'.format(num_classes), s)
    # Set number of classes num_classes.
    s = re.sub('iou_threshold: [0-9].[0-9]+',
               'iou_threshold: {}'.format(iou_threshold), s)
    
    f.write(s)

# #Have a look at the config file with various settings
!cat {pipeline_fname}

"""# Train the model

##[Optional] The cell below adds Tensorboard visualization to the training process.
Will open in new tab.
"""

#After running this cell click on the link in the output cell to open tensorboard
#Tensoarboard will show you graphically different training parameters as the model is training
#when training finishes after the set number of steps, tensorboard can be used to see a nice summary of the training process
#Visuals will load in Tensorboard after the model has gone through a few hundred steps

model_dir = '/content/models/research/training'
!wget https://bin.equinox.io/c/4VmDzA7iaHb/ngrok-stable-linux-amd64.zip
!unzip ngrok-stable-linux-amd64.zip
LOG_DIR = model_dir
get_ipython().system_raw(
    'tensorboard --logdir {} --host 0.0.0.0 --port 6006 &'
    .format(LOG_DIR)
)
get_ipython().system_raw('./ngrok http 6006 &')
! curl -s http://localhost:4040/api/tunnels | python3 -c \
    "import sys, json; print(json.load(sys.stdin)['tunnels'][0]['public_url'])"

"""## Start the training"""

model_dir = 'training/'
# Optionally remove content in output model directory for a fresh start.
!rm -rf {model_dir}
os.makedirs(model_dir, exist_ok=True)
!python /content/models/research/object_detection/model_main.py \
    --pipeline_config_path={pipeline_fname} \
    --model_dir={model_dir} \
    --alsologtostderr \
    --num_train_steps={num_steps} \
    --num_eval_steps={num_eval_steps}

#model dir check for the trained model
!ls {model_dir}

"""## Export a Trained Inference Graph
Once your training job is complete, you need to extract the newly trained inference graph, which will be later used to perform the object detection. This can be done as follows:
"""

#clean output_directory if necessary to start fresh:

# !rm -rf /content/object_detection_demo/fine_tuned_model/ 
# os.makedirs('/content/object_detection_demo_flow/fine_tuned_model/', exist_ok=True)

# Commented out IPython magic to ensure Python compatibility.
# %%capture
# import re
# import numpy as np
# 
# output_directory = './fine_tuned_model'
# # output_directory = '/content/gdrive/My\ Drive/data/'
# 
# lst = os.listdir(model_dir)
# lst = [l for l in lst if 'model.ckpt-' in l and '.meta' in l]
# steps=np.array([int(re.findall('\d+', l)[0]) for l in lst])
# last_model = lst[steps.argmax()].replace('.meta', '')
# 
# last_model_path = os.path.join(model_dir, last_model)
# print(last_model_path)
# !python /content/models/research/object_detection/export_inference_graph.py \
#     --input_type=image_tensor \
#     --pipeline_config_path={pipeline_fname} \
#     --output_directory={output_directory} \
#     --trained_checkpoint_prefix={last_model_path}

#export directory check
!ls {output_directory}

import os
pb_fname = os.path.join(os.path.abspath(output_directory), "frozen_inference_graph.pb")
assert os.path.isfile(pb_fname), '`{}` not exist'.format(pb_fname)
# !ls -alh {pb_fname}

"""## Running Inference: Checking what the trained model can detect
Test with images in repository `object_detection_demo_flow/data/images/final test` directory.
"""

import os
import glob

# Path to frozen detection graph. This is the actual model that is used for the object detection.
PATH_TO_CKPT = pb_fname

# List of the strings that is used to add correct label for each box.
PATH_TO_LABELS = label_map_pbtxt_fname

# If you want to test the code with your images, just add images files to the PATH_TO_TEST_IMAGES_DIR.
PATH_TO_TEST_IMAGES_DIR =  os.path.join(repo_dir_path, "data/images/final_test")

assert os.path.isfile(pb_fname)
assert os.path.isfile(PATH_TO_LABELS)
TEST_IMAGE_PATHS = glob.glob(os.path.join(PATH_TO_TEST_IMAGES_DIR, "*.*"))
assert len(TEST_IMAGE_PATHS) > 0, 'No image found in `{}`.'.format(PATH_TO_TEST_IMAGES_DIR)
print(TEST_IMAGE_PATHS)

# Commented out IPython magic to ensure Python compatibility.
# %cd /content/models/research/object_detection

import numpy as np
import os
import six.moves.urllib as urllib
import sys
import tarfile
import tensorflow as tf
import zipfile

from collections import defaultdict
from io import StringIO
# This is needed to display the images.
# %matplotlib inline
import matplotlib.pyplot as plt
from PIL import Image

# This is needed since the notebook is stored in the object_detection folder.
sys.path.append("..")
from object_detection.utils import ops as utils_ops

from object_detection.utils import label_map_util

from object_detection.utils import visualization_utils as vis_util


detection_graph = tf.Graph()
with detection_graph.as_default():
    od_graph_def = tf.GraphDef()
    with tf.gfile.GFile(PATH_TO_CKPT, 'rb') as fid:
        serialized_graph = fid.read()
        od_graph_def.ParseFromString(serialized_graph)
        tf.import_graph_def(od_graph_def, name='')


label_map = label_map_util.load_labelmap(PATH_TO_LABELS)
categories = label_map_util.convert_label_map_to_categories(
    label_map, max_num_classes=num_classes, use_display_name=True)
category_index = label_map_util.create_category_index(categories)


def load_image_into_numpy_array(image):
    (im_width, im_height) = image.size
    return np.array(image.getdata()).reshape(
        (im_height, im_width, 3)).astype(np.uint8)

# Size, in inches, of the output images.
IMAGE_SIZE = (12, 8)


def run_inference_for_single_image(image, graph):
    with graph.as_default():
        with tf.Session() as sess:
            # Get handles to input and output tensors
            ops = tf.get_default_graph().get_operations()
            all_tensor_names = {
                output.name for op in ops for output in op.outputs}
            tensor_dict = {}
            for key in [
                'num_detections', 'detection_boxes', 'detection_scores',
                'detection_classes', 'detection_masks'
            ]:
                tensor_name = key + ':0'
                if tensor_name in all_tensor_names:
                    tensor_dict[key] = tf.get_default_graph().get_tensor_by_name(
                        tensor_name)
            if 'detection_masks' in tensor_dict:
                # The following processing is only for single image
                detection_boxes = tf.squeeze(
                    tensor_dict['detection_boxes'], [0])
                detection_masks = tf.squeeze(
                    tensor_dict['detection_masks'], [0])
                # Reframe is required to translate mask from box coordinates to image coordinates and fit the image size.
                real_num_detection = tf.cast(
                    tensor_dict['num_detections'][0], tf.int32)
                detection_boxes = tf.slice(detection_boxes, [0, 0], [
                                           real_num_detection, -1])
                detection_masks = tf.slice(detection_masks, [0, 0, 0], [
                                           real_num_detection, -1, -1])
                detection_masks_reframed = utils_ops.reframe_box_masks_to_image_masks(
                    detection_masks, detection_boxes, image.shape[0], image.shape[1])
                detection_masks_reframed = tf.cast(
                    tf.greater(detection_masks_reframed, 0.5), tf.uint8)
                # Follow the convention by adding back the batch dimension
                tensor_dict['detection_masks'] = tf.expand_dims(
                    detection_masks_reframed, 0)
            image_tensor = tf.get_default_graph().get_tensor_by_name('image_tensor:0')

            # Run inference
            output_dict = sess.run(tensor_dict,
                                   feed_dict={image_tensor: np.expand_dims(image, 0)})

            # all outputs are float32 numpy arrays, so convert types as appropriate
            output_dict['num_detections'] = int(
                output_dict['num_detections'][0])
            output_dict['detection_classes'] = output_dict[
                'detection_classes'][0].astype(np.uint8)
            output_dict['detection_boxes'] = output_dict['detection_boxes'][0]
            output_dict['detection_scores'] = output_dict['detection_scores'][0]
            if 'detection_masks' in output_dict:
                output_dict['detection_masks'] = output_dict['detection_masks'][0]
    return output_dict


for image_path in TEST_IMAGE_PATHS:
    image = Image.open(image_path)
    print(image_path)
    # the array based representation of the image will be used later in order to prepare the
    # result image with boxes and labels on it.
    image_np = load_image_into_numpy_array(image)
    # Expand dimensions since the model expects images to have shape: [1, None, None, 3]
    image_np_expanded = np.expand_dims(image_np, axis=0)
    # Actual detection.
    output_dict = run_inference_for_single_image(image_np, detection_graph)
    # Visualization of the results of a detection.
    vis_util.visualize_boxes_and_labels_on_image_array(
        image_np,
        output_dict['detection_boxes'],
        output_dict['detection_classes'],
        output_dict['detection_scores'],
        category_index,
        instance_masks=output_dict.get('detection_masks'),
        use_normalized_coordinates=True,
        line_thickness=8)
    plt.figure(figsize=IMAGE_SIZE)
    plt.imshow(image_np)
    plt.show()

"""# Convert TF model to OpenVINO 21.03 Intermediate Representation (IR)
 This can be used to run inference on OpenVINO.
# In order to run the model on DepthAI modules, we then compile the IR obtained above to a .blob (via a server we set up just for that)

## First, we install Open Vino 21.03
"""

# Commented out IPython magic to ensure Python compatibility.
# %cd ../..

# Commented out IPython magic to ensure Python compatibility.
import os
from urllib.parse import urlparse

## install tools. Open Vino takes some time to download - it's ~400MB
!sudo apt-get install -y pciutils cpio
!sudo apt autoremove

## downnload installation files
url = "https://registrationcenter-download.intel.com/akdlm/irc_nas/17662/l_openvino_toolkit_p_2021.3.394.tgz"
!wget {url}

## Get the name of the tgz
parsed = urlparse(url)
openvino_tgz = os.path.basename(parsed.path)
openvino_folder = os.path.splitext(openvino_tgz)[0]

## Extract & install openvino
!tar xf {openvino_tgz}
# %cd {openvino_folder}
!./install_openvino_dependencies.sh && \
    sed -i 's/decline/accept/g' silent.cfg && \
    ./install.sh --silent silent.cfg

!ls

"""[Optional] Open Vino install check, generally not needed"""

# !source /opt/intel/openvino/bin/setupvars.sh && \
#     /opt/intel/openvino/deployment_tools/demo/demo_squeezenet_download_convert_run.sh

"""### Here we run some modifications in the ssd2 OpenVINO extension for TF so that our Mobilenet SSDv2 model can convert successfully to the IR"""

# Commented out IPython magic to ensure Python compatibility.
# %cd /opt/intel/openvino_2021/deployment_tools/model_optimizer/extensions/front/tf/

#openvino fixes: edit 
# Read in the file, make sure the .json corresponds to the model!!!
with open('ssd_v2_support.json', 'r') as file :
  filedata = file.read()

# Replace the target string
filedata = filedata.replace('"Postprocessor/ToFloat"', '"Postprocessor/Cast_1"')

# Write the file out again
with open('ssd_v2_support.json', 'w') as file:
  file.write(filedata)

"""## Convert TF model to Open Vino Intermediate Representation
If using own model, please change to your desired name for output directory --output_dir "choose name"
"""

# Commented out IPython magic to ensure Python compatibility.
#CONVERT TF MODEL to OPEN VINO IRv10. saved in IR_V10_fruits_mnssdv2_6k directory or
#choose own name for --output_dir "choose name"
# %cd "/content/models/research/fine_tuned_model/"
!source /opt/intel/openvino_2021/bin/setupvars.sh && \
    python /opt/intel/openvino_2021/deployment_tools/model_optimizer/mo.py \
    --input_model frozen_inference_graph.pb \
    --tensorflow_use_custom_operations_config /opt/intel/openvino_2021/deployment_tools/model_optimizer/extensions/front/tf/ssd_v2_support.json \
    --tensorflow_object_detection_api_pipeline_config pipeline.config \
    --reverse_input_channels \
    --output_dir ./fruits_10k \
    --data_type FP16

# from google.colab import drive
# drive.mount('/content/drive')

# Commented out IPython magic to ensure Python compatibility.
#check directory containing the exported TF trained model and the IRv10 folder
# %ls fruits_10k/

"""## Now we compile the IR model to a .blob for use on DepthAI modules/platform

---

### We save the blob in the IR directory from above, corresponding to --output_dir parameter above. 
The blob filename will be *frozen_inference_graph.blob*
"""

#No changes needed here unless using custom data.
#CHOOSE the directory where you would like to save the blob.
# I use the same --output_dir as above for the IR conversion
blob_dir = "/content/models/research/fine_tuned_model/fruits_10k/"

#Copy the path of your .xml and .bin files. For that, you can look at the IR
#conversion output cell, select and copy from:
#[SUCCESS] XML file and bin file paths.
#Or you can choose to compile other .xml .bin files from a different location
#
xmlfile = "/content/models/research/fine_tuned_model/fruits_10k/frozen_inference_graph.xml"
binfile = "/content/models/research/fine_tuned_model/fruits_10k/frozen_inference_graph.bin"
!python -m pip install blobconverter

import blobconverter
blob_path = blobconverter.from_openvino(
    xml=xmlfile,
    bin=binfile,
    data_type="FP16",
    shaves=5,
)
from google.colab import files
files.download(blob_path)

"""##Now you can download your .blob file and run it on the DepthAI module/platform
To download locally, use the file explorer on the left to locate the file in the --output_dir folder, then right click download. Colab takes a few seconds to prepare the file, then the download prompt will appear.

# Run the blob on DepthAI

To use this blob with DepthAI:

- Clone depthai demo script

    ```
    $ git clone https://github.com/luxonis/depthai.git
    ```
- Install requirements

    ```
    $ python3 install_requirements.py
    ```
- Create a new model directory and insert the blob

    ```
    $ mkdir resources/nn/custom_mobilenet
    $ mv <downloaded_blob> resources/nn/custom_mobilenet/
    ```

- Copy config file from mobilenet-ssd model

    ```
    $ cp resources/nn/mobilenet-ssd/mobilenet-ssd.json resources/nn/custom_mobilenet/custom_mobilenet.json
    ```

- Run the demo script

    ```
    $ python3 depthai_demo.py -cnn custom_mobilenet
    ```

You can read more details on deployment [here](https://docs.luxonis.com/en/latest/pages/tutorials/first_steps/#using-custom-models)
"""

#Compress the folder with the Tensorflow model and OpenVINO IR_V10 folder and download it locally.
#These are in content/models/research/fine_tuned_model, so we compress the fine_tuned_model folder.
!tar czvf fine_tuned_model.tar.gz {blob_dir}

#download the compressed IRv10 folder locally
#or can use file navigator on the left to move it to your gdrive
from google.colab import files
files.download("fine_tuned_model.tar.gz")

Standalone JPEG Lux-ESp32

Python
import cv2
import numpy as np
import depthai as dai
import sys
from time import sleep

'''
Basic demo of gen2 pipeline builder functionality where output of jpeg encoded images are sent out SPI rather than the typical XLink out interface.
Make sure you have something to handle the SPI protocol on the other end! See the included ESP32 example. 
'''

def create_spi_demo_pipeline():
    print("Creating SPI pipeline: ")
    print("COLOR CAM -> ENCODER -> SPI OUT")
    pipeline = dai.Pipeline()

    cam_color         = pipeline.createColorCamera()
    spiout_preview    = pipeline.createSPIOut()
    videnc            = pipeline.createVideoEncoder()

    # set up color camera and link to NN node
    cam_color.setPreviewSize(300, 300);
    cam_color.setResolution(dai.ColorCameraProperties.SensorResolution.THE_1080_P);
    cam_color.setInterleaved(False);
    cam_color.setCamId(0);
    cam_color.setColorOrder(dai.ColorCameraProperties.ColorOrder.BGR);

    # VideoEncoder
    videnc.setDefaultProfilePreset(1920, 1080, 30, dai.VideoEncoderProperties.Profile.MJPEG);

    # Link plugins CAM -> ENCODER -> SPI OUT
    cam_color.video.link(videnc.input);
    spiout_preview.setStreamName("spipreview");
    spiout_preview.setBusId(0);
    videnc.bitstream.link(spiout_preview.input);

    return pipeline

def flash_bootloader():
    (f, bl) = dai.DeviceBootloader.getFirstAvailableDevice()
    bootloader = dai.DeviceBootloader(bl)
    print(bootloader.getVersion())

    progress = lambda p : print(f'Flashing progress: {p*100:.1f}%')
    bootloader.flashBootloader(progress)


def flash_image():
    pipeline = create_spi_demo_pipeline()
    
    (found, bl) = dai.DeviceBootloader.getFirstAvailableDevice()

    if(found):
        bootloader = dai.DeviceBootloader(bl)
        progress = lambda p : print(f'Flashing progress: {p*100:.1f}%')
        bootloader.flash(progress, pipeline)
    else:
        print("No booted (bootloader) devices found...")

def write_image_to_file(filename):
    pipeline = create_spi_demo_pipeline()
    dai.DeviceBootloader.saveDepthaiApplicationPackage(filename, pipeline)

if(len(sys.argv) >= 2 and sys.argv[1] == "bootloader"):
    print("flashing bootloader")
    flash_bootloader()
elif(len(sys.argv) >= 2 and sys.argv[1] == "save"):
    filename = "pipeline.dap"
    print("saving pipeline to disk as " + filename)
    write_image_to_file(filename)
else:
    print("flashing pipeline")
    flash_image()

Image preview server

C/C++
/* HTTP File Server Example
   This example code is in the Public Domain (or CC0 licensed, at your option.)
   Unless required by applicable law or agreed to in writing, this
   software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
   CONDITIONS OF ANY KIND, either express or implied.
*/

#include <sys/param.h>

#include "esp_wifi.h"
#include "esp_event.h"
#include "esp_log.h"
#include "esp_system.h"
#include "esp_spiffs.h"
#include "nvs_flash.h"
#include "esp_netif.h"
#include "protocol_examples_common.h"





//-------------------------------------------------------------------------------------------------------------------
#include <main.hpp>

#include "esp32_spi_impl.h"
#include "spi_api.hpp"

#define MAX_DETECTIONS 16

static const char* METASTREAM = "spimetaout";
static const char* PREVIEWSTREAM = "spipreview";

const char *base_path = "/spiffs";
const char *jpeg_path = "/spiffs/frame.jpg";

extern "C" {
   void app_main();
}

// example callback for chunking up large messages.
uint32_t example_chunk_recv_size = 0;
FILE* JPEG_FILE = NULL;
bool write_jpeg = false;

uint8_t skip_print = 0;
void example_chunk_message(char* received_packet, uint32_t packet_size, uint32_t message_size){

    // if write file var is set and it's the first packet of a message...
    if( write_jpeg && !JPEG_FILE && example_chunk_recv_size == 0){
        // open up the jpeg file for writing.
        printf("starting new jpeg write...\n");
        JPEG_FILE = fopen(jpeg_path, "w");

        if (!JPEG_FILE) {
            printf("Failed to create file : %s", jpeg_path);
        }
    }   

    example_chunk_recv_size += packet_size;
    if(skip_print%16 == 0){
        printf("example_chunk_message called back packet %d/%d\n", example_chunk_recv_size, message_size);
        skip_print = 0;
    }
    skip_print++;

    // open up the jpeg file for writing.
    if(write_jpeg && JPEG_FILE){
        fwrite(received_packet, sizeof(char), packet_size, JPEG_FILE);
    }

    // if it's the last packet of a message...
    if(example_chunk_recv_size >= message_size){
        example_chunk_recv_size = 0;
        if (write_jpeg && JPEG_FILE){
            // close up the jpeg file.
            fclose(JPEG_FILE);
            JPEG_FILE = NULL;
            write_jpeg = false;
        }
    }
}

void start_depthai_spi(){
    uint8_t req_success = 0;

    dai::SpiApi mySpiApi;
    mySpiApi.set_send_spi_impl(&esp32_send_spi);
    mySpiApi.set_recv_spi_impl(&esp32_recv_spi);

    while(1) {
        // ----------------------------------------
        // example of getting large messages a chunk/packet at a time.
        // ----------------------------------------
        mySpiApi.set_chunk_packet_cb(&example_chunk_message);
        mySpiApi.chunk_message(PREVIEWSTREAM);

        // ----------------------------------------
        // pop current message/metadata. this tells the depthai to update the info being passed back using the spi_cmds.
        // ----------------------------------------
        req_success = mySpiApi.spi_pop_messages();
    }
}

extern "C" {
   void app_main();
}
//-------------------------------------------------------------------------------------------------------------------








/* This example demonstrates how to create file server
 * using esp_http_server. This file has only startup code.
 * Look in file_server.c for the implementation */

static const char *TAG="example";

/* Function to initialize SPIFFS */
static esp_err_t init_spiffs(void)
{
    ESP_LOGI(TAG, "Initializing SPIFFS");

    esp_vfs_spiffs_conf_t conf = {
      .base_path = "/spiffs",
      .partition_label = NULL,
      .max_files = 5,   // This decides the maximum number of files that can be created on the storage
      .format_if_mount_failed = true
    };

    esp_err_t ret = esp_vfs_spiffs_register(&conf);
    if (ret != ESP_OK) {
        if (ret == ESP_FAIL) {
            ESP_LOGE(TAG, "Failed to mount or format filesystem");
        } else if (ret == ESP_ERR_NOT_FOUND) {
            ESP_LOGE(TAG, "Failed to find SPIFFS partition");
        } else {
            ESP_LOGE(TAG, "Failed to initialize SPIFFS (%s)", esp_err_to_name(ret));
        }
        return ESP_FAIL;
    }

    size_t total = 0, used = 0;
    ret = esp_spiffs_info(NULL, &total, &used);
    if (ret != ESP_OK) {
        ESP_LOGE(TAG, "Failed to get SPIFFS partition information (%s)", esp_err_to_name(ret));
        return ESP_FAIL;
    }

    ESP_LOGI(TAG, "Partition size: total: %d, used: %d", total, used);
    return ESP_OK;
}

/* Declare the function which starts the file server.
 * Implementation of this function is to be found in
 * file_server.c */
esp_err_t start_file_server(const char *base_path);

void app_main(void)
{
    ESP_ERROR_CHECK(nvs_flash_init());
    ESP_ERROR_CHECK(esp_netif_init());
    ESP_ERROR_CHECK(esp_event_loop_create_default());

    /* This helper function configures Wi-Fi or Ethernet, as selected in menuconfig.
     * Read "Establishing Wi-Fi or Ethernet Connection" section in
     * examples/protocols/README.md for more information about this function.
     */
    ESP_ERROR_CHECK(example_connect());

    /* Initialize file storage */
    ESP_ERROR_CHECK(init_spiffs());

    /* Start the file server */
    ESP_ERROR_CHECK(start_file_server(base_path));



//----------------------------------------------------------------------------------------------
    // init spi for the esp32
    init_esp32_spi();

    start_depthai_spi();

    //Never reached.
    deinit_esp32_spi();
//----------------------------------------------------------------------------------------------


}

Credits

Sumit Kumar

Sumit Kumar

32 projects • 94 followers
19 y/o. My daily routine involves dealing with electronics, code, distributed storage and cloud APIs.

Comments

Related channels and tags