search

matterport / Mask_RCNN

Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow
Other
24.52k stars 11.68k forks source link

train_shapes.ipynb detects lots of nonsense #2839

Open ChrisDAT20 opened 2 years ago

ChrisDAT20 commented 2 years ago

I cloned the TF 2 branch and adapted the configuration because i have a GPU card with 2GB of memory only.

    GPU_COUNT = 1
    IMAGES_PER_GPU = 1
    USE_MINI_MASK = False
    BACKBONE = "resnet50"

This is the test image in inference mode:

grafik

And this is the result:

grafik

Is there something i missed?

tqamarVT commented 2 years ago

I believe this may have something to do with the load_weights function in model.py.

Try to forcefully reinstall h5py, restart your environment/machine, and see if you get the same result after inference.

That was the issue that was causing this same problem for me.

avinash-218 commented 1 year ago

The ‘model.load_weights’ seem to load the weights incorrectly due to version compatibility issues, resulting in training from scratch. So during training and evaluation, the coco and the previously trained weights were not loaded properly and hence in case of training, the training happens from scratch and in case of evaluation the loaded model predicts worse on the sample data.

Because of this reason, the losses at earlier steps at earlier epochs were too high and also the visual results looked random, not even close to the ground truth, and also the evaluation metrics such as mAP, mAR, F1 were 0. This can be solved by two ways : By using ‘tf.keras.Model.load_weights’ instead of ‘model.load_weights’ - But still this can’t be used since it doesn’t support the ‘exclude’ argument.

By downgrading tensorflow from 2.7 to 2.5 worked in both training (from coco using exclude argument and from previously trained weights) and also in evaluations.

This worked for me. Correct me if i am wrong or my understanding is wrong

TianRuoyu commented 5 months ago

Hi! did u solve this problem？

avinash-218 commented 2 months ago

Hi! did u solve this problem？

Yess