10:16:56 - INFO - Initializing logo detector with embedding model: models/logo_detection/clip_finetuned 10:16:56 - INFO - Initializing DetectLogosDETR on device: cuda:0 10:16:56 - INFO - DETR model: No local model found, will download from HuggingFace: Pravallika6/detr-finetuned-logo-detection_v2 10:16:56 - INFO - Loading DETR model: Pravallika6/detr-finetuned-logo-detection_v2 Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.52, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`. Device set to use cuda:0 10:16:58 - INFO - Embedding model: No local model found, will download from HuggingFace: models/logo_detection/clip_finetuned 10:16:58 - INFO - Loading fine-tuned CLIP model from: models/logo_detection/clip_finetuned 10:16:58 - INFO - Base model: openai/clip-vit-large-patch14 Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.52, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`. 10:17:00 - INFO - Fine-tuned CLIP model loaded successfully 10:17:00 - INFO - DetectLogosDETR initialization complete 10:17:00 - INFO - Loading ground truth from database... 10:17:01 - INFO - Loaded ground truth for 158654 test images 10:17:01 - INFO - Sampling 50 reference logos with 3 refs each... 10:17:01 - INFO - Selected 50 reference logos 10:17:01 - INFO - Computing reference logo embeddings... Reference logos: 0%| | 0/50 [00:00