Visualize model predictions

[mzhadigerov]

I ran the pre-trained model in eval mode and got this output:

python main.py --mode eval --data_type structure --config_file structure_config.json --data_root_dir data/ --model_load_path data/model/structure.pth --debug --device cpu
{'lr': 5e-05, 'lr_backbone': 1e-05, 'batch_size': 2, 'weight_decay': 0.0001, 'epochs': 20, 'lr_drop': 1, 'lr_gamma': 0.9, 'clip_max_norm': 0.1, 'backbone': 'resnet18', 'num_classes': 6, 'dilation': False, 'position_embedding': 'sine', 'emphasized_weights': {}, 'enc_layers': 6, 'dec_layers': 6, 'dim_feedforward': 2048, 'hidden_dim': 256, 'dropout': 0.1, 'nheads': 8, 'num_queries': 125, 'pre_norm': True, 'masks': False, 'aux_loss': False, 'mask_loss_coef': 1, 'dice_loss_coef': 1, 'ce_loss_coef': 1, 'bbox_loss_coef': 5, 'giou_loss_coef': 2, 'eos_coef': 0.4, 'set_cost_class': 1, 'set_cost_bbox': 5, 'set_cost_giou': 2, 'device': 'cpu', 'seed': 42, 'start_epoch': 0, 'num_workers': 2, 'data_root_dir': 'data/', 'config_file': 'structure_config.json', 'data_type': 'structure', 'model_load_path': 'data/model/structure.pth', 'metrics_save_filepath': '', 'table_words_dir': None, 'mode': 'eval', 'debug': True, 'checkpoint_freq': 1, '__module__': '__main__', '__dict__': <attribute '__dict__' of 'Args' objects>, '__weakref__': <attribute '__weakref__' of 'Args' objects>, '__doc__': None}
----------------------------------------------------------------------------------------------------
loading model
loading model from checkpoint
loading data
creating index...
index created!
Test:  [0/1]  eta: 0:00:00  class_error: 0.00  loss: 0.3392 (0.3392)  loss_ce: 0.0231 (0.0231)  loss_bbox: 0.0250 (0.0250)  loss_giou: 0.2912 (0.2912)  loss_ce_unscaled: 0.0231 (0.0231)  class_error_unscaled: 0.0000 (0.0000)  loss_bbox_unscaled: 0.0050 (0.0050)  loss_giou_unscaled: 0.1456 (0.1456)  cardinality_error_unscaled: 0.0000 (0.0000)  time: 0.3716  data: 0.0614  max mem: 0
Test: Total time: 0:00:00 (0.3762 s / it)
Averaged stats: class_error: 0.00  loss: 0.3392 (0.3392)  loss_ce: 0.0231 (0.0231)  loss_bbox: 0.0250 (0.0250)  loss_giou: 0.2912 (0.2912)  loss_ce_unscaled: 0.0231 (0.0231)  class_error_unscaled: 0.0000 (0.0000)  loss_bbox_unscaled: 0.0050 (0.0050)  loss_giou_unscaled: 0.1456 (0.1456)  cardinality_error_unscaled: 0.0000 (0.0000)
Accumulating evaluation results...
DONE (t=0.01s).
IoU metric: bbox
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.619
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.750
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.629
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.619
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.281
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.506
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.638
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.638
pubmed: AP50: 0.750, AP75: 0.629, AP: 0.619, AR: 0.638

How can I visualize the model predictions on input images? like : this

[bsmock]

We just pushed an update to the code and documentation related to this. Please take a look at the updated documentation and how to use the --debug flag during evaluation to visualize results.

It should produce a visualization of the (overlapping) bounding boxes directly output by the model:

...and of the final table cells after post-processing:

Cheers, Brandon

[mzhadigerov]

Thanks @bsmock ! But how do I make inference If all I have is an image (without json file containing texts and coordinates required for evaluation) ? Basically all I want to do is to test the pre-trained table structure recognizer model on my jpg image containing a table and visualize the recognized table components (rows, headers, columns etc.).

[mzhadigerov]

@hannody Could you attach the original image, please? I'm getting ValueError: not supported on this line:

outputs = self.model(img_tensor)

Here is my file:

infer_v2.zip

[hannody]

Hi @mzhadigerov, try this one please table Then compare your results with mine, I have attached the structure before and after the post-processing, but the post-processing is not complete

[achillesliu]

@mzhadigerov in the predict function from TableStructure class, tensor shape manipulating process has some logic errors. It runs through but leads to your ValueError: not supported error

[mzhadigerov]

@hannody Thanks, I was able to solve the bug. Regarding your images (before/after postprocessing): so the secod image (with red rectangles) is postprocessed? The first one looks more legit to me.

[hannody]

@hannody Thanks, I was able to solve the bug. Regarding your images (before/after postprocessing): so the secod image (with red rectangles) is postprocessed? The first one looks more legit to me.

My bad, sorry for the confusion, the one with only red rects is the raw one, uses the output data directly, not sure how you viewed the notebook, but this should be obvious inside the notebook, let me know if you improved upon it and good luck.

[mzhadigerov]

@hannody yes, that's what I thought it was according to the code in the notebook. Thanks for clarification!

[sontakke12297]

@hannody Could you attach the original image, please? I'm getting ValueError: not supported on this line:

outputs = self.model(img_tensor)

Here is my file:

infer_v2.zip

@mzhadigerov How did you solve this issue?