Thank you

I would be interested to see how that goes.

I hear what you're saying and I will try to take that on board for the next video.

There is not much in the way of pins for the esp32-cam. You can probably push it to about three or four extra pins if you use pin 16, 1, 3 and 0. But you need pin 1 and 3 for flashing (while 0 is held low). If you do OTA updates, you can probably work out something with those pins or you can work out some kind of pin-sharing arrangement? My advice would be to use a port expander or another esp32 dev board with an esp32-now communication (I have tried the latter on another project with good results).

Hi. I didn't end up using the OSD because it didn't work with the colour image I used the rgb_print function which is from github.com/MakeMagazinDE/ESP32-CAM/blob/master/app_httpd.cpp.

@@jonathanr4242 Thank you. I assume that print function is injected into the video stream? So you can only see those values if you have video?

@@madladdan No. The opposite. I only overlay on a still picture. So, I have to convert it to rgb888, print out to the image, and then convert back to jpeg and save to memeory. The code is on my github repository. I also print out to the tft screen (same text), but this won't show up on the image file, only the tft screen. Also, you need to have esp32 board version 1.0.6 installed for this to work. //print to image rgb888_matrix = dl_matrix3du_alloc(1, fb->width, fb->height, 3); fmt2rgb888(fb->buf, fb->len, fb->format, rgb888_matrix->item); // HERE print some text // rgb_print(rgb888_matrix, 0x000000FF, "Hello!"); for (size_t ix = 0; ix < EI_CLASSIFIER_LABEL_COUNT; ix++) { char tmp[32]; sprintf(tmp, "%s: %0.2f", result.classification[ix].label, result.classification[ix].value); rgb_print(rgb888_matrix, 0x000000FF, tmp, ix*15 + 2); } Serial.println("printing to image"); size_t _jpg_buf_len = 0; uint8_t * _jpg_buf = NULL; // free(fb->buf); // fb->buf = NULL; bool jpeg_converted = fmt2jpg(rgb888_matrix->item, fb->width*fb->height*3, fb->width, fb->height, PIXFORMAT_RGB888, 80, &_jpg_buf, &_jpg_buf_len); //&fb->buf Serial.println("converted jpg");

@@jonathanr4242 Oh I see. Thank you

That would be dependent on what device the usb camera is connected to. The camera is just the capture mechanism but the image needs to be processed on some kind of computer chip.

@@jonathanr4242 could you suggest which type of computer chip we can use to get the picture apart from computer.

@@47lokeshkumar74 Ok. You mean a usb camera connected to a pc. There are a number of options for deployment. When you get to the deployment step, there is an option to "run your impulse directly" on a computer. I would look into that.

The RobotZero website, referenced in the description, has the schematic for connecting the esp32-cam to the tft screen. The only other thing I use is a push button, which is on pin 4 of the esp32 cam. If you are using TFT_eSPI library you need to copy the User_Setup.h file from my github repo (the first github link in the description).

😆🤣

That's a very good idea The issue is, the neural network classification runs a bit slow on the esp32 for a dash cam. We could use motion detection so it records images when it detects movement.

@@jonathanr4242 but actually i dont need it detects anything just for recording image while you driving car ~~