Accuracy Checker reshape error in yolov5

[Ashwin_J_S]

Hi, I have a custom trained yolov5 model, When i convert to fp32 and run accuracy checker on it i am facing a reshape error. to verify my configuration yaml file i have tried the same configuration file in pretrained yolov5 fp32 model and the accuracy checker works fine there . Is there anyways to merge two channels in openvino ?? Have attached the error screen grab

The yaml configuration file i used 

models:
  - name: yolo_v5
    launchers:
      - framework: dlsdk
        model: /home/nga_hitech_ib/yolov5_MBU/yolov5s.xml
        weights: /home/nga_hitech_ib/yolov5_MBU/yolov5s.bin
        device: CPU
        adapter:
          type: yolo_v5
          anchors: "10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326"
          num: 3
          coords: 4
          classes: 80
          threshold: 0.001
          anchor_masks: [[0, 1, 2], [3, 4, 5], [6, 7, 8]]
          raw_output: True
          outputs:
            - '345'
            - '404'
            - '463'
           


    datasets:
      - name: COCO2017_detection_80cl
        data_source: /home/nga_hitech_ib/model_test/coco_2017/val2017
        annotation_conversion:
            converter : mscoco_detection
            annotation_file: /home/nga_hitech_ib/model_test/coco_2017/instances_val2017.json
            images_dir: /home/nga_hitech_ib/model_test/coco_2017/val2017
            has_background: false
            use_full_label_map: false

        preprocessing:
          - type: resize
            size: 640
        postprocessing:
          - type: resize_prediction_boxes
          - type: filter
            apply_to: prediction
            min_confidence: 0.001
            remove_filtered: true
          - type: nms
            overlap: 0.5
          - type: clip_boxes
            apply_to: prediction
        metrics:
          - type: map
            integral: 11point
            ignore_difficult: true
            presenter: print_scalar
          - name: AP@0.5
            type: coco_precision
            max_detections: 100
            threshold: 0.5
          - name: AP@0.5:0.05:95
            type: coco_precision
            max_detections: 100
            threshold: '0.5:0.05:0.95'

[Wan_Intel]

Hi Ashwin_J_S,

Thanks for your information.

I’ll share the information with next level for further investigation. Meanwhile, is it possible to share your custom model with us?

Regards,

Wan

[Ashwin_J_S]

Thank you for the update @Wan_Intel , But as I mentioned before unfortunately I am prohibited to share the model with you. 

[Wan_Intel]

Hi Ashwin_J_S,

Noted with thanks.

Could you please share the config file you used with the Accuracy Checker?

Regards,

Wan

[Ashwin_J_S]

Sure @Wan_Intel ,

models:
  - name: yolo_v5
    launchers:
      - framework: dlsdk
        device: CPU
        adapter:
          type: yolo_v5
          anchors: "10,13, 16,30, 33,23, 30,61, 62,45, 59,119, 116,90, 156,198, 373,326"
          num: 3
          coords: 4
          classes: 80
          threshold: 0.001
          anchor_masks: [[0, 1, 2], [3, 4, 5], [6, 7, 8]]
          raw_output: True
          outputs:
            - '443'
            - '487'
            - '531'
            
            

    datasets:
      - name: small
        data_source: "/home/ubuntu/coco_2017/val2017/"
        annotation_conversion:
          converter: mscoco_detection
          annotation_file: "/home/ubuntu/coco_2017/annotations/instances_val2017.json"
          images_dir: "/home/ubuntu/coco_2017/val2017/"

        preprocessing:
          - type: resize
            size: 640

        postprocessing:
          - type: resize_prediction_boxes
          - type: filter
            apply_to: prediction
            min_confidence: 0.001
            remove_filtered: True
          - type: nms
            overlap: 0.5
          - type: clip_boxes
            apply_to: prediction

        metrics:
          - type: map
            integral: 11point
            ignore_difficult: true
            presenter: print_scalar
          - type: coco_precision
            max_detections: 100
            threshold: 0.5

[Wan_Intel]

Hi Ashwin_J_S,

Could you please use anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]] in the config file and run Accuracy Checker again?

Regards,

Wan

[Wan_Intel]

Hi Ashwin_J_S,

Are you able to use your custom model with Benchmark Python Tool?

Regards,

Wan

[Ashwin_J_S]

Hi @Wan_Intel , Yes I am able to use the benchmark_app  with my custom model.

benchmark_app -m last.xml -niter 100 

[Wan_Intel]

Hi Ashwin_J_S,

Could you share the command you use to run Accuracy Checker?

Regards,

Wan

[Ashwin_J_S]

Yeah sure  @Wan_Intel ,

accuracy_check -m last.xml -c acc.yaml 

[Wan_Intel]

Hi Ashwin_J_S,

@Ashwin_J_S 

I encountered the same error when I change the classes number.

May I know how many classes did you used to train the model?

Regards,

Wan

[Ashwin_J_S · Accepted solution]

Hi, @Wan_Intel  

So I remembered I used only 4 classes , and changed 80 to 4 in the configuration file its working now, Thank you so much for help.

[Wan_Intel]

Hi Ashwin_J_S,

@Ashwin_J_S 

Great! We’re happy that you are now able to run your model with Accuracy Checker.

This thread will no longer be monitored since this issue has been resolved. 

If you need any additional information from Intel, please submit a new question.

Thanks, and best regards,

Wan