config.json
| 1 | { |
| 2 | "architectures": [ |
| 3 | "YolosForObjectDetection" |
| 4 | ], |
| 5 | "attention_probs_dropout_prob": 0.0, |
| 6 | "auxiliary_loss": false, |
| 7 | "bbox_cost": 5, |
| 8 | "bbox_loss_coefficient": 5, |
| 9 | "class_cost": 1, |
| 10 | "eos_coefficient": 0.1, |
| 11 | "giou_cost": 2, |
| 12 | "giou_loss_coefficient": 2, |
| 13 | "hidden_act": "gelu", |
| 14 | "hidden_dropout_prob": 0.0, |
| 15 | "hidden_size": 384, |
| 16 | "id2label": { |
| 17 | "0": "N/A", |
| 18 | "1": "person", |
| 19 | "2": "bicycle", |
| 20 | "3": "car", |
| 21 | "4": "motorcycle", |
| 22 | "5": "airplane", |
| 23 | "6": "bus", |
| 24 | "7": "train", |
| 25 | "8": "truck", |
| 26 | "9": "boat", |
| 27 | "10": "traffic light", |
| 28 | "11": "fire hydrant", |
| 29 | "12": "N/A", |
| 30 | "13": "stop sign", |
| 31 | "14": "parking meter", |
| 32 | "15": "bench", |
| 33 | "16": "bird", |
| 34 | "17": "cat", |
| 35 | "18": "dog", |
| 36 | "19": "horse", |
| 37 | "20": "sheep", |
| 38 | "21": "cow", |
| 39 | "22": "elephant", |
| 40 | "23": "bear", |
| 41 | "24": "zebra", |
| 42 | "25": "giraffe", |
| 43 | "26": "N/A", |
| 44 | "27": "backpack", |
| 45 | "28": "umbrella", |
| 46 | "29": "N/A", |
| 47 | "30": "N/A", |
| 48 | "31": "handbag", |
| 49 | "32": "tie", |
| 50 | "33": "suitcase", |
| 51 | "34": "frisbee", |
| 52 | "35": "skis", |
| 53 | "36": "snowboard", |
| 54 | "37": "sports ball", |
| 55 | "38": "kite", |
| 56 | "39": "baseball bat", |
| 57 | "40": "baseball glove", |
| 58 | "41": "skateboard", |
| 59 | "42": "surfboard", |
| 60 | "43": "tennis racket", |
| 61 | "44": "bottle", |
| 62 | "45": "N/A", |
| 63 | "46": "wine glass", |
| 64 | "47": "cup", |
| 65 | "48": "fork", |
| 66 | "49": "knife", |
| 67 | "50": "spoon", |
| 68 | "51": "bowl", |
| 69 | "52": "banana", |
| 70 | "53": "apple", |
| 71 | "54": "sandwich", |
| 72 | "55": "orange", |
| 73 | "56": "broccoli", |
| 74 | "57": "carrot", |
| 75 | "58": "hot dog", |
| 76 | "59": "pizza", |
| 77 | "60": "donut", |
| 78 | "61": "cake", |
| 79 | "62": "chair", |
| 80 | "63": "couch", |
| 81 | "64": "potted plant", |
| 82 | "65": "bed", |
| 83 | "66": "N/A", |
| 84 | "67": "dining table", |
| 85 | "68": "N/A", |
| 86 | "69": "N/A", |
| 87 | "70": "toilet", |
| 88 | "71": "N/A", |
| 89 | "72": "tv", |
| 90 | "73": "laptop", |
| 91 | "74": "mouse", |
| 92 | "75": "remote", |
| 93 | "76": "keyboard", |
| 94 | "77": "cell phone", |
| 95 | "78": "microwave", |
| 96 | "79": "oven", |
| 97 | "80": "toaster", |
| 98 | "81": "sink", |
| 99 | "82": "refrigerator", |
| 100 | "83": "N/A", |
| 101 | "84": "book", |
| 102 | "85": "clock", |
| 103 | "86": "vase", |
| 104 | "87": "scissors", |
| 105 | "88": "teddy bear", |
| 106 | "89": "hair drier", |
| 107 | "90": "toothbrush" |
| 108 | }, |
| 109 | "image_size": [ |
| 110 | 512, |
| 111 | 864 |
| 112 | ], |
| 113 | "initializer_range": 0.02, |
| 114 | "intermediate_size": 1536, |
| 115 | "label2id": { |
| 116 | "N/A": 83, |
| 117 | "airplane": 5, |
| 118 | "apple": 53, |
| 119 | "backpack": 27, |
| 120 | "banana": 52, |
| 121 | "baseball bat": 39, |
| 122 | "baseball glove": 40, |
| 123 | "bear": 23, |
| 124 | "bed": 65, |
| 125 | "bench": 15, |
| 126 | "bicycle": 2, |
| 127 | "bird": 16, |
| 128 | "boat": 9, |
| 129 | "book": 84, |
| 130 | "bottle": 44, |
| 131 | "bowl": 51, |
| 132 | "broccoli": 56, |
| 133 | "bus": 6, |
| 134 | "cake": 61, |
| 135 | "car": 3, |
| 136 | "carrot": 57, |
| 137 | "cat": 17, |
| 138 | "cell phone": 77, |
| 139 | "chair": 62, |
| 140 | "clock": 85, |
| 141 | "couch": 63, |
| 142 | "cow": 21, |
| 143 | "cup": 47, |
| 144 | "dining table": 67, |
| 145 | "dog": 18, |
| 146 | "donut": 60, |
| 147 | "elephant": 22, |
| 148 | "fire hydrant": 11, |
| 149 | "fork": 48, |
| 150 | "frisbee": 34, |
| 151 | "giraffe": 25, |
| 152 | "hair drier": 89, |
| 153 | "handbag": 31, |
| 154 | "horse": 19, |
| 155 | "hot dog": 58, |
| 156 | "keyboard": 76, |
| 157 | "kite": 38, |
| 158 | "knife": 49, |
| 159 | "laptop": 73, |
| 160 | "microwave": 78, |
| 161 | "motorcycle": 4, |
| 162 | "mouse": 74, |
| 163 | "orange": 55, |
| 164 | "oven": 79, |
| 165 | "parking meter": 14, |
| 166 | "person": 1, |
| 167 | "pizza": 59, |
| 168 | "potted plant": 64, |
| 169 | "refrigerator": 82, |
| 170 | "remote": 75, |
| 171 | "sandwich": 54, |
| 172 | "scissors": 87, |
| 173 | "sheep": 20, |
| 174 | "sink": 81, |
| 175 | "skateboard": 41, |
| 176 | "skis": 35, |
| 177 | "snowboard": 36, |
| 178 | "spoon": 50, |
| 179 | "sports ball": 37, |
| 180 | "stop sign": 13, |
| 181 | "suitcase": 33, |
| 182 | "surfboard": 42, |
| 183 | "teddy bear": 88, |
| 184 | "tennis racket": 43, |
| 185 | "tie": 32, |
| 186 | "toaster": 80, |
| 187 | "toilet": 70, |
| 188 | "toothbrush": 90, |
| 189 | "traffic light": 10, |
| 190 | "train": 7, |
| 191 | "truck": 8, |
| 192 | "tv": 72, |
| 193 | "umbrella": 28, |
| 194 | "vase": 86, |
| 195 | "wine glass": 46, |
| 196 | "zebra": 24 |
| 197 | }, |
| 198 | "layer_norm_eps": 1e-12, |
| 199 | "model_type": "yolos", |
| 200 | "num_attention_heads": 6, |
| 201 | "num_channels": 3, |
| 202 | "num_detection_tokens": 100, |
| 203 | "num_hidden_layers": 12, |
| 204 | "patch_size": 16, |
| 205 | "qkv_bias": true, |
| 206 | "torch_dtype": "float32", |
| 207 | "transformers_version": "4.19.0.dev0", |
| 208 | "use_mid_position_embeddings": true |
| 209 | } |
| 210 | |