Darknet based yolov3 training and Inference in Tensorflow for single class (Kangaroo)
I referred and used these fantastic repos during the implementation:
1. Dataset preparation and data augmentation. I have used free Roboflow for format conversion and augmentation
learning_rate=0.001 max_batches = 6000 steps=4000,5400
There are two steps:
2. Testing on sample dataset taken from Kaggle sample_data
#Some results: