5.8 Increase the warm up bias to 0.101
Case 7. Increase the warmup bias to 0.101
  1. lr0 -> 0.1
  2. momentum ->0.92
  3. lrf ->0.2
  4. weight_decay: 0.000499
  5. warmup_epochs -> 2.5
  6. warmup –momentum -> 0.801
  7. warmup-bias -> 0.101
  8. optimizer : SGD
In this experiment, we revert back to our best weight decay and increase the warmup bias learning rate to 0.101. The result obtained was PR AUC 0.128 at mAP@0.5 | F1 max 0.09 at 0.116 confidence which successfully was an improvement to the model. Figure 15 shows the F1-Confidence curve which depicts the score of F1 max 0.09 at 0.116 confidence rate which is better as compared to the result achieved. Figure 16 shows the Precision- Recall curve which depicts the score of PR AUC 0.128 at mAP@0.5 which is better as compared to the result achieved. The result obtained was PR AUC 0.128 at mAP@0.5 | F1 max 0.09 at 0.116 confidence which successfully was an improvement to the model.