config.json

{
  "_name_or_path": "google/owlvit-base-patch32",
  "architectures": [
    "OwlViTForObjectDetection"
  ],
  "id2label": {
    "0": "Coverall",
    "1": "Face_Shield",
    "2": "Gloves",
    "3": "Goggles",
    "4": "Mask"
  },
  "initializer_factor": 1.0,
  "label2id": {
    "Coverall": 0,
    "Face_Shield": 1,
    "Gloves": 2,
    "Goggles": 3,
    "Mask": 4
  },
  "logit_scale_init_value": 2.6592,
  "model_type": "owlvit",
  "projection_dim": 512,
  "text_config": {
    "bos_token_id": 0,
    "dropout": 0.0,
    "eos_token_id": 2,
    "max_length": 16,
    "model_type": "owlvit_text_model",
    "pad_token_id": 1
  },
  "text_config_dict": null,
  "torch_dtype": "float32",
  "transformers_version": "4.41.2",
  "vision_config": {
    "dropout": 0.0,
    "model_type": "owlvit_vision_model"
  },
  "vision_config_dict": null
}