Returns traffic sign with about 94% accuracy given an image.

See https://www.kaggle.com/code/dima806/traffic-sign-detection-vit for more details.

image/png

Classification report:

                              precision    recall  f1-score   support

           Bicycles crossing     1.0000    0.9660    0.9827       206
           Children crossing     0.8583    1.0000    0.9238       206
                Danger Ahead     0.9810    1.0000    0.9904       206
 Dangerous curve to the left     0.7981    0.8293    0.8134       205
Dangerous curve to the right     0.8182    0.7902    0.8040       205
                Dont Go Left     1.0000    0.9903    0.9951       206
       Dont Go Left or Right     1.0000    1.0000    1.0000       206
               Dont Go Right     1.0000    0.9610    0.9801       205
            Dont Go straight     1.0000    1.0000    1.0000       205
    Dont Go straight or left     0.9493    1.0000    0.9740       206
     Dont overtake from Left     0.9533    0.9903    0.9714       206
                      Fences     0.9762    1.0000    0.9880       205
                     Go Left     0.9844    0.9175    0.9497       206
            Go Left or right     0.8723    1.0000    0.9318       205
                    Go Right     1.0000    0.9854    0.9926       205
         Go left or straight     0.7946    0.8683    0.8298       205
        Go right or straight     0.8920    0.7621    0.8220       206
                 Go straight     0.9624    0.8689    0.9133       206
        Go straight or right     1.0000    0.8010    0.8895       206
     Heavy Vehicle Accidents     0.9579    1.0000    0.9785       205
                        Horn     1.0000    1.0000    1.0000       206
                      No Car     1.0000    1.0000    1.0000       205
                    No Uturn     0.9856    1.0000    0.9928       206
                    No entry     1.0000    1.0000    1.0000       205
                     No horn     1.0000    1.0000    1.0000       205
                 No stopping     0.9856    1.0000    0.9927       205
                Road Divider     1.0000    1.0000    1.0000       206
        Roundabout mandatory     0.9951    1.0000    0.9976       205
        Speed limit (15km/h)     1.0000    1.0000    1.0000       206
        Speed limit (30km/h)     0.9619    0.9806    0.9712       206
        Speed limit (40km/h)     0.9800    0.9515    0.9655       206
        Speed limit (50km/h)     0.9757    0.9757    0.9757       206
         Speed limit (5km/h)     1.0000    0.9951    0.9976       206
        Speed limit (60km/h)     1.0000    0.4126    0.5842       206
        Speed limit (70km/h)     1.0000    0.9466    0.9726       206
              Train Crossing     0.9671    1.0000    0.9833       206
          Under Construction     1.0000    0.9806    0.9902       206
                     Unknown     1.0000    0.5415    0.7025       205
                       Uturn     1.0000    1.0000    1.0000       205
              Zebra Crossing     0.9206    0.9563    0.9381       206
                ZigZag Curve     0.8047    1.0000    0.8918       206
                   keep Left     0.7895    0.8010    0.7952       206
                  keep Right     0.8565    0.9902    0.9186       205
        speed limit (80km/h)     0.6042    0.9854    0.7491       206
          watch out for cars     1.0000    1.0000    1.0000       205

                    accuracy                         0.9388      9252
                   macro avg     0.9472    0.9388    0.9366      9252
                weighted avg     0.9472    0.9388    0.9366      9252
Downloads last month
67
Safetensors
Model size
85.8M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for dima806/traffic_sign_detection

Finetuned
(1769)
this model