Open e76971072 opened 1 year ago
because for no reason, your Test dataset and training dataset should have same number of class in theirs cocodataset file... For me test missing one class ( 5) compare 6 for training, I added a picture about this classes in test and all it was correct after
order = _validate_interpolation_order(image.dtype, order) 100/100 [==============================] - ETA: 0s - batch: 49.5000 - size: 4.0000 - loss: 2.1849 - rpn_class_loss: 0.1415 - rpn_bbox_loss: 0.6310 - mrcnn_class_loss: 0.1085 - mrcnn_bbox_loss: 0.6098 - mrcnn_mask_loss: 0.6940
ValueError Traceback (most recent call last) in
----> 1 vis_img.train_model(num_epochs = 100, augmentation=True,path_trained_models = "/content/drive/MyDrive/Background-Image-Processor-Reseller/BackGround-Image-Processor-Reseller/Beanie/models")
9 frames /usr/local/lib/python3.8/dist-packages/keras/engine/training_utils_v1.py in standardize_input_data(data, names, shapes, check_batch_axis, exception_prefix) 640 for dim, ref_dim in zip(data_shape, shape): 641 if ref_dim != dim and ref_dim is not None and dim is not None: --> 642 raise ValueError('Error when checking ' + exception_prefix + 643 ': expected ' + names[i] + ' to have shape ' + 644 str(shape) + ' but got array with shape ' +
ValueError: Error when checking input: expected input_image_meta to have shape (18,) but got array with shape (16,)
Code snippet
vis_img = instance_custom_training()
vis_img.modelConfig(network_backbone = "resnet101", num_classes= 5, batch_size = 4)
vis_img.load_pretrained_model("mask_rcnn_coco.h5")
vis_img.load_dataset("./Beanie")
vis_img.train_model(num_epochs = 100, augmentation=True,path_trained_models = "models")
Can you anyone help me resolve this error ? somehow it couldn't save the weights during last epoch run.