{ "_name_or_path": "facebook/detr-resnet-50", "activation_dropout": 0.0, "activation_function": "relu", "architectures": [ "DetrForObjectDetection" ], "attention_dropout": 0.0, "auxiliary_loss": false, "backbone": "resnet50", "backbone_config": null, "backbone_kwargs": { "in_chans": 3, "out_indices": [ 1, 2, 3, 4 ] }, "bbox_cost": 5, "bbox_loss_coefficient": 5, "class_cost": 1, "classifier_dropout": 0.0, "d_model": 256, "decoder_attention_heads": 8, "decoder_ffn_dim": 2048, "decoder_layerdrop": 0.0, "decoder_layers": 6, "dice_loss_coefficient": 1, "dilation": false, "dropout": 0.1, "encoder_attention_heads": 8, "encoder_ffn_dim": 2048, "encoder_layerdrop": 0.0, "encoder_layers": 6, "eos_coefficient": 0.1, "giou_cost": 2, "giou_loss_coefficient": 2, "id2label": { "1": 63, "2": 63, "3": 76, "4": 77, "5": 106, "6": 107, "7": 2, "8": 7, "9": 28, "10": 29, "11": 3, "12": 59, "13": 60, "14": 32, "15": 87, "16": 89, "17": 1, "18": 70, "19": 14, "20": 6, "21": 63, "22": 83, "23": 7, "24": 9, "25": 66, "26": 20, "27": 19, "28": 67, "29": 101, "30": 100, "31": 102, "32": 92, "33": 82, "34": 23, "35": 24, "36": 17, "37": 22, "38": 20, "39": 19, "40": 63, "41": 64, "42": 8, "43": 87, "44": 57, "45": 58, "46": 51, "47": 98, "48": 99, "49": 44, "50": 45, "51": 46, "52": 31, "53": 35, "54": 34, "55": 50, "56": 97, "57": 98, "58": 99, "59": 69, "60": 67, "61": 84, "62": 95, "63": 96, "64": 74, "65": 81, "66": 80, "67": 78, "68": 79, "69": 70, "70": 71, "71": 42, "72": 43, "73": 57, "74": 58, "75": 97, "76": 98, "77": 99, "78": 32, "79": 33, "80": 49, "81": 15, "82": 12, "83": 51, "84": 74, "85": 75, "86": 53, "87": 94, "88": 92, "89": 106, "90": 107, "91": 2, "92": 85, "93": 86, "94": 20, "95": 19, "96": 26, "97": 72, "98": 103, "99": 104, "100": 105, "101": 72, "102": 73, "103": 74, "104": 18, "105": 61, "106": 62, "107": 10, "108": 87, "109": 88, "110": 61, "111": 62, "112": 47, "113": 48, "114": 11, "115": 13, "116": 25, "117": 23, "118": 24, "119": 92, "120": 93, "121": 35, "122": 34, "123": 83, "124": 36, "125": 37, "126": 38, "127": 39, "128": 40, "129": 52, "130": 53, "131": 15, "132": 12, "133": 4, "134": 101, "135": 100, "136": 102, "137": 90, "138": 91, "139": 10, "140": 8, "141": 28, "142": 30, "143": 54, "144": 55, "145": 56, "146": 53, "147": 14, "148": 16, "149": 81, "150": 80, "151": 78, "152": 41, "153": 42, "154": 21, "155": 68, "156": 28, "157": 29, "158": 26, "159": 27 }, "init_std": 0.02, "init_xavier_std": 1.0, "is_encoder_decoder": true, "label2id": { "1": 17, "2": 91, "3": 11, "4": 133, "6": 20, "7": 23, "8": 140, "9": 24, "10": 139, "11": 114, "12": 132, "13": 115, "14": 147, "15": 131, "16": 148, "17": 36, "18": 104, "19": 95, "20": 94, "21": 154, "22": 37, "23": 117, "24": 118, "25": 116, "26": 158, "27": 159, "28": 156, "29": 157, "30": 142, "31": 52, "32": 78, "33": 79, "34": 122, "35": 121, "36": 124, "37": 125, "38": 126, "39": 127, "40": 128, "41": 152, "42": 153, "43": 72, "44": 49, "45": 50, "46": 51, "47": 112, "48": 113, "49": 80, "50": 55, "51": 83, "52": 129, "53": 146, "54": 143, "55": 144, "56": 145, "57": 73, "58": 74, "59": 12, "60": 13, "61": 110, "62": 111, "63": 40, "64": 41, "66": 25, "67": 60, "68": 155, "69": 59, "70": 69, "71": 70, "72": 101, "73": 102, "74": 103, "75": 85, "76": 3, "77": 4, "78": 151, "79": 68, "80": 150, "81": 149, "82": 33, "83": 123, "84": 61, "85": 92, "86": 93, "87": 108, "88": 109, "89": 16, "90": 137, "91": 138, "92": 119, "93": 120, "94": 87, "95": 62, "96": 63, "97": 75, "98": 76, "99": 77, "100": 135, "101": 134, "102": 136, "103": 98, "104": 99, "105": 100, "106": 89, "107": 90 }, "mask_loss_coefficient": 1, "max_position_embeddings": 1024, "model_type": "detr", "num_channels": 3, "num_hidden_layers": 6, "num_queries": 100, "position_embedding_type": "sine", "scale_embedding": false, "torch_dtype": "float32", "transformers_version": "4.41.1", "use_pretrained_backbone": true, "use_timm_backbone": true }