Bad mAP results after long train

[HomirSumpson1]

Im training the model with resnet152 backbone for pothole detection and im getting bad results. My dataset has 3k images where 2k are for training and the rest for validation, i have 4 classes of potholes and after 7 hours of training with 2k steps (20 epochs since the training crashes because of the datasetGenerator) the model is getting only 0.21 mAP. Im new to this so can you guys give some tips to improve the mAP of the model?

[lucysumi]

Hi, I'm also getting the same GeneratorDataset error and training stops around 15-17 epochs. Did you find the solution for this?
Any help would be appreciated. TYIA!