Object-Detection

This project is used for object detection using Yolov3 and Fast RCNN.

YOLO and Fast RCNN are two different types of object detectors.

• Yolo is a Single Shot detectors.
• Fast RCNN is a Region based deetctors.

Demo

YOLOv3	Fast-RCNN

• Yolov3 seems to have more accuracy than Fast Rcnn.
• Yolov3 execution Frames per second is far better than Fast RCNN.
• Yolov3 seems to detect small object precisely than larger object.
• Fast RCNN works great on high resolution pictures and is more accurate.
• For real time usage i think Yolov3 is better to implement in current scenario.(Although Faster RCNN can be used. we will try to cover it in future.)

Download the necessary weights and files.

Fast RCNN model

wget http://download.tensorflow.org/models/object_detection/mask_rcnn_inception_v2_coco_2018_01_28.tar.gz tar zxvf mask_rcnn_inception_v2_coco_2018_01_28.tar.gz

YOLOv3

wget https://pjreddie.com/media/files/yolov3.weights

wget https://github.com/pjreddie/darknet/blob/master/cfg/yolov3.cfg?raw=true -O ./yolov3.cfg

Dependencies

• Opencv

Usage

• data folder contains all the input and output files.
• Run mask_rcnn.py and yolo.py for Fast RCNN and Yolov3 respectively.
• mscoco_labels.names contain names of object which can be detected.

Commandline usage for object detection using YOLOv3 and Fast RCNN.

a single image:

python yolo.py --image=cars.jpg

python mask_rcnn.py --image=data\cars.jpg

a video file:

python yolo.py --video=run.mp4

python3 mask_rcnn.py --video=run.mp4

If no argrument is passed, it start's the webcam.

Please feel free to raise any issues, if anything is wrong in your concern.

Credit

https://medium.com/@jonathan_hui/what-do-we-learn-from-single-shot-object-detectors-ssd-yolo-fpn-focal-loss-3888677c5f4d

https://medium.com/@jonathan_hui/object-detection-speed-and-accuracy-comparison-faster-r-cnn-r-fcn-ssd-and-yolo-5425656ae359

https://github.com/spmallick