How to convert the PyTorch Data Loader to sv.Detections Object

[shanalikhan]

Search before asking

• I have searched the Supervision issues and found no similar feature requests.

Question

I've Dataloader object and I am able to convert the predicted model output using from_transformer to Detection Object. Since i want to generate the MAP and confusion matrix, I need to convert the actual object to Detection object as well.

How to do that?

for idx,batch in enumerate(tqdm(TEST_DATALOADER)):
    
    pixel_values = batch["pixel_values"].to(DEVICE)
    pixel_mask = batch["pixel_mask"].to(DEVICE)
    #print(batch["labels"])
    labels = [{k: v.to(DEVICE) for k, v in t.items()} for t in batch["labels"]]
    #print(labels)
    with torch.no_grad(): 
      outputs = model(pixel_values=pixel_values, pixel_mask=pixel_mask)
      #outputs = model(**batch)

    orig_target_sizes = torch.stack([target["orig_size"] for target in labels], dim=0)
    #print(outputs)
    results = image_processor.post_process_object_detection(outputs, target_sizes=orig_target_sizes,threshold=0.5)
    print("LABELS")
    print(labels[0].keys())
    
    print(labels[0]['class_labels'])
    print("RESULTS")
    print(results[0].keys())
    print(results[0]['labels'])
    for result in results:
      detections = sv.Detections.from_transformers(result)#.with_nms(threshold=IOU_TRESHOLD)
      #sv.Detections.from_
      targets.append(annotations)
      predictions.append(detections)
    break

Additional

No response