Realtime Multi-object Detection Pipeline

Note: this repo is currently under heavy development. It's not ready for general consumption. So, please refrain yourself from using it in production.

The goal of this project is to buid a single end-to-end deep learning model for more accurate and faster (near real-time) multi-object detection that can be train in single-pass of multiple different pieces:

• Single Shot MultiBox Detector (SSD)

• YOLOv3 real-time properties

• Focal loss for dense object detection (RetinaNet)

• Non Maximum Suppression (NMS)

• Scalable object detection using deep neural networks

• Faster R-CNN tricks

These techniques and methods from various research papers will be implemented using PyTorch.

We will be using Pascal VOC2007 dataset.

Requirements

• Python 3

• Pytorch 0.4

• numpy

• fastai PyTorch library

Training

# Select the script that you want to train for reproducing a results./retina_ce_sgd_0.001.sh# For the focal loss use ./retina_focal_sgd_0.0001.sh

You can see the details in trainer.py

VOC Dataset

Download VOC2007 trainval & test

# specify a directory for dataset to be downloaded into, else default is ~/data/sh data/scripts/VOC2007.sh # <directory>

Download VOC2012 trainval

# specify a directory for dataset to be downloaded into, else default is ~/data/sh data/scripts/VOC2012.sh # <directory>

TODO

•  Build SSD + YOLO model

•  Apply cross entropy loss and focal loss

•  Compare between CE loss and focal loss

•  Report results on VOC

• currently achieved 50mAP on VOC2007.

•  Report results on COCO

•  Use relative path for easy reproducing of result