Yolov3 is an algorithm that uses convolutional neural network to perform object detection. Open source two endpoints each for image and video detections which can be integrate in any application. An example of integration with iOS app is shown in this repo.
- Two Endpoints (image and Video)
- Bounding boxes
- 80 objects can be classified and detected.
.
├── doc # Canvas documents, presentations
├── src # Source files
├── pubs # Documentation files
│ ├── figs # Figures
└── README.md
Tensorflow CPU
pip install -r requirements.txt
yolov3 For Windows:
You can download weights by clicking here here
Now you have to load the models and convert them to tensorflow checkpoint files.
yolov3
python load_weights.py
python app.py
This will start a web server on localhost. To check the detection open the Postman, select body > form-data > file and upload image.
endpoints
http://localhost:5000/image
http://localhost:5000/detections
Detections API (http://localhost:5000/detections)
This route takes in the image input and returns a JSON response with all the detections found within each image (classes found within the images and the associated confidence)
Image API (http://localhost:5000/image)
This route takes in the image input and returns the image response with all the detections found within the image and their bounding boxes.
Video can only be run through this command so far.
python detect_video.py --video path-to-video
To run the iOS application run the .xcodeproj found in src/AIes-Ios./Swift. In order for the application to run, the application must be in lanscape mode and must have the URL in line #778 updated according to the service running.
- Akash Sindhu
- Maninder Singh
- Tommy Chivaz
- Jiawei Zhang