Detection Updated

[GOSHROW]

I have added configuration files for yolov3 as was told to. The weights could not be added for yolov3 as they exceeded file size. I have tried to add weights for yolov3tiny however it returned 403 access denied during push. I also tried up the NFPA dataset, however, the weights file is too large for the yolov3 architecture. furthermore,every epoch is computationally heavy for yolov3.

To avoid such failure in approach, I did in issue #5 , I added the approach through SSD-mobile net on coco and other datasets. these changes have been directly made on app.py than on detection. The automation for switching to SSD than yolov3 is left open for the indexer algorithm. However, SSD has provided more reliability in practical uses. The SSD related prototext and weight-config files may be seen in the repository: dataset.

The wieghts and cfg for yolov3 have been added to yolo-coco.

[ayulockin]

Nice PR. I am reviewing this. Will let reach out to you for queries.

[ayulockin]

Can you list down the major changes in the PR after your detailed introduction?

[GOSHROW]

The major changes are :

• SSD detection has been added as an alternative to the YOLOv3.
• Prototxt and CaffeModels have been added for the SSD.
• The switching and control-flow for detection models has not been added.
• YOLO config files have been added to train upon.
• YOLO coco pre-trained weights and cfg have been added.
• pbtxt and config files have been added for SSD