Skip to content

Latest commit

 

History

History
6 lines (4 loc) · 409 Bytes

File metadata and controls

6 lines (4 loc) · 409 Bytes

Deep-Learning-Techniques-To-Predict-Number-Of-Objects-In-Image-

  • Developed NN for fve way classifcation of 10k,16×16 RGBA images using softmax activation and cross entropy loss
  • Improved F1-score by varying hidden layer size and depth, and employing mini-batch SGD with adaptive learning rates
  • Achieved a maximum accuracy of 76% across fve diferently shaped objects using sigmoid and ReLU activation