the model I made is over 500mb which exceeds github's max file requirements. I cannot provide my training files either as they exceed 500mb.
This is my implementation of object detection using OpenCV and python. The model I trained is specifically trained to recognize animals. Hoping that some non-profit organizations can use this in the future so categorize and identify endangered species (I know a non-profit that does this for the endangered kenya giraffe and thats how I got the idea). Depending on CPU, takes about .5 seconds to recognize each image so non-profits could use at a rate of 120 images/minute
OpenCV dnn module supports running inference on pre-trained deep learning models from popular frameworks like Caffe, Torch and TensorFlow.
When it comes to object detection, popular detection frameworks are
- YOLO
- SSD
- Faster R-CNN
Support for running YOLO/DarkNet has been added to OpenCV dnn module recently.
- opencv
- numpy
pip install numpy opencv-python
Note: Compatability with Python 2.x is not officially tested.
Download the pre-trained YOLO v3 weights file from this link and place it in the current directory or you can directly download to the current directory in terminal using
$ wget https://pjreddie.com/media/files/yolov3.weights
Provided all the files are in the current directory, below command will apply object detection on the input image dog.jpg.
$ python yolo_opencv.py --image dog.jpg --config yolov3.cfg --weights yolov3.weights --classes yolov3.txt
Command format
$ python yolo_opencv.py --image /path/to/input/image --config /path/to/config/file --weights /path/to/weights/file --classes /path/to/classes/file
Checkout the blog post to learn more.
Checkout the object detection implementation available in cvlib which enables detecting common objects in the context through a single function call detect_common_objects().
(SSD and Faster R-CNN examples will be added soon)