AttributeError: 'NoneType' object has no attribute 'copy' about detectron HOT 1 CLOSED

There are two problems:

1. You are trying to use an ImageNet pretrained model as the source of model weights, but this model was never trained for keypoint prediction. You need to change it to the weights from a keypoints model file.

2. The test-time augmentation yaml config is actually only set up for inference with precomputed RPN proposals on the COCO minival set.

Here's an example for keypoint inference with an end-to-end trained keypoints Mask R-CNN model:

python2 tools/infer_simple.py \
    --cfg configs/12_2017_baselines/e2e_keypoint_rcnn_R-50-FPN_s1x.yaml \
    --output-dir /tmp/detectron-visualizations \
    --image-ext jpg \
    --wts https://s3-us-west-2.amazonaws.com/detectron/37697714/12_2017_baselines/e2e_keypoint_rcnn_R-50-FPN_s1x.yaml.08_44_03.qrQ0ph6M/output/train/keypoints_coco_2014_train:keypoints_coco_2014_valminusminival/generalized_rcnn/model_final.pkl \
    demo

from detectron.