This is an online image search platform running on the Raspberry Pi via Object detection through Tensorflow lite.
- Search image via URL link
- Search image via upload an image
- Search similar images
- numpy==1.16.2
- Flask>=1.0.0
- requests==2.22.0
- Pillow>=6.2.0
- Google Edge TPU or Tensorflow Lite (Follow the instructions here)
It uses the flask in python, just one command to run the web server simply:
$ python3 server.py
Then visit the website: 127.0.0.1:5000 (port 5000 on the Raspberry Pi).
Here‘s screenshots of running on the Raspberry Pi 3B+ with Coral USB Accelerator:
-
Coral USB Accelerator (Optional)
Important Things：This program does not have to use the Coral USB Accelerator, but I used it for accelerating the inference process in Tensorflow lite.
Thanks for object detection source code in the coral example, this image search platform is based on this. I used the MobileNet SSD v2 (COCO) model by default, you can modify it in the file detect_image.py .
If you don't have a usb accelerator, you can compile the entire Tensorflow lite on the Raspberry Pi and learn about its object detection here.