search

davide-coccomini / MINTIME-Multi-Identity-size-iNvariant-TIMEsformer-for-Video-Deepfake-Detection

Code for Video Deepfake Detector from "MINTIME: Multi-Identity Size-Invariant Video Deepfake Detection", paper available on IEEE Transactions on Information Forensics and Security.
58 stars 9 forks source link

video run error #9

Open pradeepkc11 opened 1 year ago

pradeepkc11 commented 1 year ago

File "/home/pradeep/MINTIME-Multi-Identity-size-iNvariant-TIMEsformer-for-Video-Deepfake-Detection/predict.py", line 555, in pred, identity_attentions, aggregated_attentions, identities, frames_per_identity = predict(opt.video_path, clustered_faces, config, opt) File "/home/pradeep/MINTIME-Multi-Identity-size-iNvariant-TIMEsformer-for-Video-Deepfake-Detection/predict.py", line 393, in predict videos, size_embeddings, mask, identities_mask, positions, tokens_per_identity = generate_masks(video_path, identities, discarded_faces, config["model"]["num-frames"], config["model"]["image-size"], config["model"]["num-patches"]) File "/home/pradeep/MINTIME-Multi-Identity-size-iNvariant-TIMEsformer-for-Video-Deepfake-Detection/predict.py", line 325, in generate_masks transformed_images = transform(image=sequence[0], image1=sequence[1], image2=sequence[2], image3=sequence[3], image4=sequence[4], image5=sequence[5], image6=sequence[6], image7=sequence[7], image8=sequence[8], image9=sequence[9], image10=sequence[10], image11=sequence[11], image12=sequence[12], image13=sequence[13], image14=sequence[14], image15=sequence[15]) File "/home/pradeep/MINTIME-Multi-Identity-size-iNvariant-TIMEsformer-for-Video-Deepfake-Detection/lib/python3.10/site-packages/albumentations/core/composition.py", line 195, in call self._check_args(**data) File "/home/pradeep/MINTIME-Multi-Identity-size-iNvariant-TIMEsformer-for-Video-Deepfake-Detection/lib/python3.10/site-packages/albumentations/core/composition.py", line 286, in _check_args raise ValueError( ValueError: Height and Width of image, mask or masks should be equal. You can disable shapes check by setting a parameter is_check_shapes=False of Compose class (do it only if you are sure about your data consistency).

davide-coccomini commented 11 months ago

Can you give me more details?