esp32-mjpeg-multiclient-espcam-drivers/PlatformIO/README.md

MJPEG Multiclient Streaming Server

THIS DOCUMENT IS WORK IN PROGRESS

With latest ESP-CAM drivers

Full story: https://www.hackster.io/anatoli-arkhipenko/multi-client-mjpeg-streaming-from-esp32-47768f

Video: https://youtu.be/bsAVJSZeSmc

Updated on 2024-09-26:

• moved to Platform IO

• code includes all three RTOS frame delivery options:

-- Single RTOS task serving the same frame to all clients (compile CAMERA_MULTICLIENT_QUEUE)

-- Separate dedicated RTOS tasks serving the same frame to clients independently (compile CAMERA_MULTICLIENT_TASK)

-- Serarate dedicated RTOS tasks serving individual frames to clients - no frames are "dropped" (compile CAMERA_ALL_FRAMES)

Quick start

Pull the repo recursively in order to include the latest esp32-camera drivers.

Notes:

What is this?

MJPEG Multiclient Streaming Server using RTOS tasks and queues to serve video frames to clients from the onboard camera.

Several video frame delivery options are available:

• Single RTOS task serving the same frame to all clients (compile CAMERA_MULTICLIENT_QUEUE)

• Separate dedicated RTOS tasks serving the same frame to clients independently (compile CAMERA_MULTICLIENT_TASK)

• Serarate dedicated RTOS tasks serving individual frames to clients - no frames are "dropped" (compile CAMERA_ALL_FRAMES)

Compile options - Camera types

The following camearas are supported. (configurable in the platformio.ini file) Only ESP32-CAM and ESP-EYE were actually tested.

Pinouts for some/all cameras are described here: https://randomnerdtutorials.com/esp32-cam-camera-pin-gpios/

• ESP-EYE (CAMERA_MODEL_ESP_EYE)

• ESP32-CAM AI-Thinker (CAMERA_MODEL_AI_THINKER)

These cameras could be supported, but require additional work (creating board and partition files, and editing of platformio.ini file)

• Freenove ESP32-Wrover (CAMERA_MODEL_WROVER_KIT)

• M5 Cameras (models A and B) (CAMERA_MODEL_M5STACK_PSRAM)

• M5-stack ESP32 camera without PSRAM (CAMERA_MODEL_M5STACK_WIDE)

• TTGO T-Journal (CAMERA_TTGO_T_JOURNAL)

• TTGO T-Camera Plus (CAMERA_TTGO_T_PLUS)

• TTGO T-Camera Plus (CAMERA_TTGO_PIR)

Compile options - Frame size

The following frame sizes are currently supported by the camera: (configurable in the platformio.ini file)

• FRAMESIZE_96X96, // 96x96

• FRAMESIZE_QQVGA, // 160x120

• FRAMESIZE_QCIF, // 176x144

• FRAMESIZE_HQVGA, // 240x176

• FRAMESIZE_240X240, // 240x240

• FRAMESIZE_QVGA, // 320x240

• FRAMESIZE_CIF, // 400x296

• FRAMESIZE_HVGA, // 480x320

• FRAMESIZE_VGA, // 640x480

• FRAMESIZE_SVGA, // 800x600

• FRAMESIZE_XGA, // 1024x768

• FRAMESIZE_HD, // 1280x720

• FRAMESIZE_SXGA, // 1280x1024

• FRAMESIZE_UXGA, // 1600x1200

Larger frames are served slower

Latest camera drivers

This repo references Espressif's latest camera drivers' git repo directly as a component. The drivers re located at: https://github.com/espressif/esp32-camera

To update to the latest - run git pull origin command from the esp32-camera folder.

DISCLAIMER:

The repository should compile and work AS-IS as of the date stated above. Espressif is actively working on the camera drivers, and their future updates may break the procedure below. Please report the broken process providing as much information as possible, definitely the make and model of your camera device, version of the Arduino Core and IDE used.

Remember, this is a hack, a POC and a test. This is NOT GUARANTEED to work on all ESP32-based devices. The performance could be different depending on the make, brand and manufacturer of your camera.

Please do not demand fixes and updates - you are welcome to take this repo as a baseline and improve upon it.

Have fun!

Procedure:

Use this process ONLY if you want to update to the very latest drivers.

Remember: updating to the latest drivers may break the code dependencies and require investigation / code changes.

1. Clone or pull this repo locally using GIT

2. Use AS-IS (in Arduino IDE) for guaranteed results. If you feel adventurous and brave - proceed to step 3.

3. Download latest ZIP file from https://github.com/espressif/esp32-camera.git into the esp32-cam subfolder

4. In the archive: delete examples and test folders

5. Delete ALL FILES in the sketch folder (from step 1) except esp32-cam*.ino and camera_pins.h

6. In the archive: switch to subfolder esp32-camera-master/target and delete subfolders for esp32s2 and esp32s3 - I have not tested with those ones

7. unzip using unzip -jo esp32-camera-master.zip command. This will place all files in the same folder in a flat file structure

   NOTE: please observe the -jo flag: the sketch assumes all files are in the same folder and will overwrite the existing old files without asking for confirmation.

In esp32-cam.ino sketch select your camera pin assignment.

The choices are:

• CAMERA_MODEL_WROVER_KIT

• CAMERA_MODEL_ESP_EYE

• CAMERA_MODEL_M5STACK_PSRAM

• CAMERA_MODEL_M5STACK_WIDE

• CAMERA_MODEL_AI_THINKER

Compile the esp32-cam.ino sketch using the following settings:

• ESP32 Dev Module
• CPU Freq: 240
• Flash Freq: 80
• Flash mode: QIO
• Flash Size: 4Mb
• Partition: Default, Minimal SPIFFS (or any other that would fit the sketch)
• PSRAM: Enabled

Results:

I was able to run multiple browser windows, multiple VLC windows and connect multiple Blynk video widgets (max: 10) to ESP-EYE chip. The delay on the browser window was almost unnoticeable. In VLC you notice a 1 second delay due to buffering. Blynk performance all depends on the phone, so no comments there.

This is incredible considering the size of this thing! The camera on ESP-EYE is actually quite good.

Enjoy!

---

Other repositories that may be of interest

ESP32 MJPEG streaming server servicing a single client:

https://github.com/arkhipenko/esp32-cam-mjpeg

ESP32 MJPEG streaming server servicing multiple clients (FreeRTOS based):

https://github.com/arkhipenko/esp32-cam-mjpeg-multiclient

ESP32 MJPEG streaming server servicing multiple clients (FreeRTOS based) with the latest camera drivers from espressif.

https://github.com/arkhipenko/esp32-mjpeg-multiclient-espcam-drivers

Cooperative multitasking library:

https://github.com/arkhipenko/TaskScheduler