Call to predict_from_dl failing when calling convert_raw_prediction

[jcin2ai]

I am tryinng yo vall a mmdet.mask_rcnn model with backbone resnet50_fpn_1x and passing the PIL Image object, byt then in the call to conver_raw_prediction I get the next error:

ERROR:    Traceback (most recent call last):
  File "/usr/local/lib/python3.9/threading.py", line 937, in _bootstrap
    self._bootstrap_inner()
  File "/usr/local/lib/python3.9/threading.py", line 980, in _bootstrap_inner
    self.run()
  File "/usr/local/lib/python3.9/site-packages/anyio/_backends/_asyncio.py", line 867, in run
    result = context.run(func, *args)
  File "/usr/src/app/main.py", line 223, in detection
    res = measures.measure('medidas', image_file)
  File "/usr/src/app/core/utils/measure_utils.py", line 110, in measure
    return self.infer_image(model, model_type, class_map, img_obj, img_path, img_size)
  File "/usr/src/app/core/utils/measure_utils.py", line 36, in infer_image
    preds = model_type.predict_from_dl(model, infer_dl, keep_images=True, detection_threshold=0.33)
  File "/usr/local/lib/python3.9/site-packages/icevision/models/mmdet/common/mask/prediction.py", line 66, in predict_from_dl
    return _predict_from_dl(
  File "/usr/local/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.9/site-packages/icevision/models/utils.py", line 107, in _predict_from_dl
    preds = predict_fn(
  File "/usr/local/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
    return func(*args, **kwargs)
  File "/usr/local/lib/python3.9/site-packages/icevision/models/mmdet/common/mask/prediction.py", line 31, in _predict_batch
    return convert_raw_predictions(
  File "/usr/local/lib/python3.9/site-packages/icevision/models/mmdet/common/mask/prediction.py", line 93, in convert_raw_predictions
    return [
  File "/usr/local/lib/python3.9/site-packages/icevision/models/mmdet/common/mask/prediction.py", line 94, in <listcomp>
    convert_raw_prediction(
  File "/usr/local/lib/python3.9/site-packages/icevision/models/mmdet/common/mask/prediction.py", line 120, in convert_raw_prediction
    keep_masks = MaskArray(np.vstack(raw_masks)[keep_mask])
  File "<__array_function__ internals>", line 200, in vstack
  File "/usr/local/lib/python3.9/site-packages/numpy/core/shape_base.py", line 296, in vstack
    return _nx.concatenate(arrs, 0, dtype=dtype, casting=casting)
  File "<__array_function__ internals>", line 200, in concatenate
ValueError: all the input arrays must have same number of dimensions, but the array at index 0 has 2 dimension(s) and the array at index 7 has 3 dimension(s)

I am using the next Dockerfile to run the code:

FROM python:3.9.16 

# Install system dependencies
RUN apt-get update && apt-get upgrade && \
    apt remove --autoremove nvidia-cuda-toolkit && \
    apt remove --autoremove nvidia-* && \
    apt-get purge nvidia* && \
    apt-get autoremove && \
    apt-get autoclean && \
    rm -rf /usr/local/cuda* && \
    apt update && \
    apt-get install -y python-tk && \
    pip3 install --upgrade setuptools pip wheel
    
RUN pip install nvidia-pyindex
RUN pip install nvidia-cuda-runtime-cu11

# Set working directory and copy application files
WORKDIR /usr/src/app
COPY ./app /usr/src/app/

# Install Python dependencies
RUN pip3 install --no-cache-dir torch==1.10.0+cu111 torchvision==0.11.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html 
RUN pip3 install mmcv-full==1.3.17 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.10.0/index.html
RUN pip3 install mmdet==2.17.0 icevision[all]
RUN pip3 install --no-cache-dir  uvicorn fastapi pyproj python-multipart sahi==0.10.1  pillow==8.4.0

# Expose port and set command to run the application
EXPOSE 8000
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]

and the image shape is:

Image /tmp/tmpv3b24q6n/image.jpg: (619, 825, 3)

[jcin2ai]

Been looking into the code and think the error might arise from the lack of this line of code in the function convert_raw_prediction on 0.12.0

empty_mask = np.full((0, sample["img"].shape[-2], sample["img"].shape[-1]), False)
filled_raw_masks = [mask if mask != [] else empty_mask for mask in raw_masks]