projecto-MMC-DTx-2022

Vehicle make and model classification from images, using state-of-the-art computer vision models.

Dataset labels

The file stanford_train.json contains a dictionary of labels to use with a custom-defined training dataset from the Stanford Cars dataset.

• The key labels contains a list of classes for all training examples, in integer format
• The keys bbox_x1, bbox_y1, bbox_x2 and bbox_y2 contain pixel coordinates of the vehicle bounding boxes, for all training examples
• The key label_refs contains a list of strings, describing each of the 196 classes. The index of each string corresponds to the integer used as a class in labels