·
4 commits
to main
since this release
Weights after the first training of the model, the task of which is to recognize road signs of the Russian Federation
The weights of the first trained model on a training sample containing 2500 images of road signs of the Russian Federation. The training consisted of 110 epochs. The best epoch is 109: losses - 0.16296, accuracy - 0.97516. Last weights take weights from the last epoch: losses - 0.16717, accuracy - 0.97498
Model predictions based on a valid dataset:
