Amazon Robotics Challenge
About the datasets
- The datasets come from UC Berkeley and are composed of 27 different things.
sources can be found here
About the model
- The Convolutional Neural Network which I choose is the VGG19 from Oxford Visual Geometry Group
- Simply, I use the weights trained on ImageNet to initialize the Network and retrain it for the ARC datasets
Result
- Until now, I trained Network under 27 classes.
- The Result given below is trained under 27 classes.
Train Accuarcy
Train Loss
Valid Accuarcy
Valid Loss
Training Time on a single GTX 1080Ti
