{ "_commit_hash": null, "activation_dropout": 0.0, "activation_function": "relu", "architectures": [ "DetaForObjectDetection" ], "assign_first_stage": true, "attention_dropout": 0.0, "auxiliary_loss": false, "backbone_config": { "_name_or_path": "", "add_cross_attention": false, "architectures": null, "attention_probs_dropout_prob": 0.0, "bad_words_ids": null, "begin_suppress_tokens": null, "bos_token_id": null, "chunk_size_feed_forward": 0, "cross_attention_hidden_size": null, "decoder_start_token_id": null, "depths": [ 2, 2, 18, 2 ], "diversity_penalty": 0.0, "do_sample": false, "drop_path_rate": 0.1, "early_stopping": false, "embed_dim": 192, "encoder_no_repeat_ngram_size": 0, "encoder_stride": 32, "eos_token_id": null, "exponential_decay_length_penalty": null, "finetuning_task": null, "forced_bos_token_id": null, "forced_eos_token_id": null, "hidden_act": "gelu", "hidden_dropout_prob": 0.0, "hidden_size": 1536, "id2label": { "0": "LABEL_0", "1": "LABEL_1" }, "image_size": 224, "initializer_range": 0.02, "is_decoder": false, "is_encoder_decoder": false, "label2id": { "LABEL_0": 0, "LABEL_1": 1 }, "layer_norm_eps": 1e-05, "length_penalty": 1.0, "max_length": 20, "min_length": 0, "mlp_ratio": 4.0, "model_type": "swin", "no_repeat_ngram_size": 0, "num_beam_groups": 1, "num_beams": 1, "num_channels": 3, "num_heads": [ 6, 12, 24, 48 ], "num_layers": 4, "num_return_sequences": 1, "out_features": [ "stage2", "stage3", "stage4" ], "output_attentions": false, "output_hidden_states": false, "output_scores": false, "pad_token_id": null, "patch_size": 4, "path_norm": true, "prefix": null, "problem_type": null, "pruned_heads": {}, "qkv_bias": true, "remove_invalid_values": false, "repetition_penalty": 1.0, "return_dict": true, "return_dict_in_generate": false, "sep_token_id": null, "stage_names": [ "stem", "stage1", "stage2", "stage3", "stage4" ], "suppress_tokens": null, "task_specific_params": null, "temperature": 1.0, "tf_legacy_loss": false, "tie_encoder_decoder": false, "tie_word_embeddings": true, "tokenizer_class": null, "top_k": 50, "top_p": 1.0, "torch_dtype": null, "torchscript": false, "transformers_version": "4.27.0.dev0", "typical_p": 1.0, "use_absolute_embeddings": false, "use_bfloat16": false, "window_size": 12 }, "bbox_cost": 5, "bbox_loss_coefficient": 5, "class_cost": 1, "d_model": 256, "decoder_attention_heads": 8, "decoder_ffn_dim": 2048, "decoder_layerdrop": 0.0, "decoder_layers": 6, "decoder_n_points": 4, "dice_loss_coefficient": 1, "dropout": 0.1, "encoder_attention_heads": 8, "encoder_ffn_dim": 2048, "encoder_layerdrop": 0.0, "encoder_layers": 6, "encoder_n_points": 4, "eos_coefficient": 0.1, "focal_alpha": 0.25, "giou_cost": 2, "giou_loss_coefficient": 2, "id2label": { "0": "None", "1": "Person", "2": "Sneakers", "3": "Chair", "4": "Other Shoes", "5": "Hat", "6": "Car", "7": "Lamp", "8": "Glasses", "9": "Bottle", "10": "Desk", "11": "Cup", "12": "Street Lights", "13": "Cabinet/shelf", "14": "Handbag/Satchel", "15": "Bracelet", "16": "Plate", "17": "Picture/Frame", "18": "Helmet", "19": "Book", "20": "Gloves", "21": "Storage box", "22": "Boat", "23": "Leather Shoes", "24": "Flower", "25": "Bench", "26": "Potted Plant", "27": "Bowl/Basin", "28": "Flag", "29": "Pillow", "30": "Boots", "31": "Vase", "32": "Microphone", "33": "Necklace", "34": "Ring", "35": "SUV", "36": "Wine Glass", "37": "Belt", "38": "Monitor/TV", "39": "Backpack", "40": "Umbrella", "41": "Traffic Light", "42": "Speaker", "43": "Watch", "44": "Tie", "45": "Trash bin Can", "46": "Slippers", "47": "Bicycle", "48": "Stool", "49": "Barrel/bucket", "50": "Van", "51": "Couch", "52": "Sandals", "53": "Basket", "54": "Drum", "55": "Pen/Pencil", "56": "Bus", "57": "Wild Bird", "58": "High Heels", "59": "Motorcycle", "60": "Guitar", "61": "Carpet", "62": "Cell Phone", "63": "Bread", "64": "Camera", "65": "Canned", "66": "Truck", "67": "Traffic cone", "68": "Cymbal", "69": "Lifesaver", "70": "Towel", "71": "Stuffed Toy", "72": "Candle", "73": "Sailboat", "74": "Laptop", "75": "Awning", "76": "Bed", "77": "Faucet", "78": "Tent", "79": "Horse", "80": "Mirror", "81": "Power outlet", "82": "Sink", "83": "Apple", "84": "Air Conditioner", "85": "Knife", "86": "Hockey Stick", "87": "Paddle", "88": "Pickup Truck", "89": "Fork", "90": "Traffic Sign", "91": "Balloon", "92": "Tripod", "93": "Dog", "94": "Spoon", "95": "Clock", "96": "Pot", "97": "Cow", "98": "Cake", "99": "Dinning Table", "100": "Sheep", "101": "Hanger", "102": "Blackboard/Whiteboard", "103": "Napkin", "104": "Other Fish", "105": "Orange/Tangerine", "106": "Toiletry", "107": "Keyboard", "108": "Tomato", "109": "Lantern", "110": "Machinery Vehicle", "111": "Fan", "112": "Green Vegetables", "113": "Banana", "114": "Baseball Glove", "115": "Airplane", "116": "Mouse", "117": "Train", "118": "Pumpkin", "119": "Soccer", "120": "Skiboard", "121": "Luggage", "122": "Nightstand", "123": "Tea pot", "124": "Telephone", "125": "Trolley", "126": "Head Phone", "127": "Sports Car", "128": "Stop Sign", "129": "Dessert", "130": "Scooter", "131": "Stroller", "132": "Crane", "133": "Remote", "134": "Refrigerator", "135": "Oven", "136": "Lemon", "137": "Duck", "138": "Baseball Bat", "139": "Surveillance Camera", "140": "Cat", "141": "Jug", "142": "Broccoli", "143": "Piano", "144": "Pizza", "145": "Elephant", "146": "Skateboard", "147": "Surfboard", "148": "Gun", "149": "Skating and Skiing shoes", "150": "Gas stove", "151": "Donut", "152": "Bow Tie", "153": "Carrot", "154": "Toilet", "155": "Kite", "156": "Strawberry", "157": "Other Balls", "158": "Shovel", "159": "Pepper", "160": "Computer Box", "161": "Toilet Paper", "162": "Cleaning Products", "163": "Chopsticks", "164": "Microwave", "165": "Pigeon", "166": "Baseball", "167": "Cutting/chopping Board", "168": "Coffee Table", "169": "Side Table", "170": "Scissors", "171": "Marker", "172": "Pie", "173": "Ladder", "174": "Snowboard", "175": "Cookies", "176": "Radiator", "177": "Fire Hydrant", "178": "Basketball", "179": "Zebra", "180": "Grape", "181": "Giraffe", "182": "Potato", "183": "Sausage", "184": "Tricycle", "185": "Violin", "186": "Egg", "187": "Fire Extinguisher", "188": "Candy", "189": "Fire Truck", "190": "Billiards", "191": "Converter", "192": "Bathtub", "193": "Wheelchair", "194": "Golf Club", "195": "Briefcase", "196": "Cucumber", "197": "Cigar/Cigarette", "198": "Paint Brush", "199": "Pear", "200": "Heavy Truck", "201": "Hamburger", "202": "Extractor", "203": "Extension Cord", "204": "Tong", "205": "Tennis Racket", "206": "Folder", "207": "American Football", "208": "earphone", "209": "Mask", "210": "Kettle", "211": "Tennis", "212": "Ship", "213": "Swing", "214": "Coffee Machine", "215": "Slide", "216": "Carriage", "217": "Onion", "218": "Green beans", "219": "Projector", "220": "Frisbee", "221": "Washing Machine/Drying Machine", "222": "Chicken", "223": "Printer", "224": "Watermelon", "225": "Saxophone", "226": "Tissue", "227": "Toothbrush", "228": "Ice cream", "229": "Hot-air balloon", "230": "Cello", "231": "French Fries", "232": "Scale", "233": "Trophy", "234": "Cabbage", "235": "Hot dog", "236": "Blender", "237": "Peach", "238": "Rice", "239": "Wallet/Purse", "240": "Volleyball", "241": "Deer", "242": "Goose", "243": "Tape", "244": "Tablet", "245": "Cosmetics", "246": "Trumpet", "247": "Pineapple", "248": "Golf Ball", "249": "Ambulance", "250": "Parking meter", "251": "Mango", "252": "Key", "253": "Hurdle", "254": "Fishing Rod", "255": "Medal", "256": "Flute", "257": "Brush", "258": "Penguin", "259": "Megaphone", "260": "Corn", "261": "Lettuce", "262": "Garlic", "263": "Swan", "264": "Helicopter", "265": "Green Onion", "266": "Sandwich", "267": "Nuts", "268": "Speed Limit Sign", "269": "Induction Cooker", "270": "Broom", "271": "Trombone", "272": "Plum", "273": "Rickshaw", "274": "Goldfish", "275": "Kiwi fruit", "276": "Router/modem", "277": "Poker Card", "278": "Toaster", "279": "Shrimp", "280": "Sushi", "281": "Cheese", "282": "Notepaper", "283": "Cherry", "284": "Pliers", "285": "CD", "286": "Pasta", "287": "Hammer", "288": "Cue", "289": "Avocado", "290": "Hamimelon", "291": "Flask", "292": "Mushroom", "293": "Screwdriver", "294": "Soap", "295": "Recorder", "296": "Bear", "297": "Eggplant", "298": "Board Eraser", "299": "Coconut", "300": "Tape Measure/Ruler", "301": "Pig", "302": "Showerhead", "303": "Globe", "304": "Chips", "305": "Steak", "306": "Crosswalk Sign", "307": "Stapler", "308": "Camel", "309": "Formula 1", "310": "Pomegranate", "311": "Dishwasher", "312": "Crab", "313": "Hoverboard", "314": "Meat ball", "315": "Rice Cooker", "316": "Tuba", "317": "Calculator", "318": "Papaya", "319": "Antelope", "320": "Parrot", "321": "Seal", "322": "Butterfly", "323": "Dumbbell", "324": "Donkey", "325": "Lion", "326": "Urinal", "327": "Dolphin", "328": "Electric Drill", "329": "Hair Dryer", "330": "Egg tart", "331": "Jellyfish", "332": "Treadmill", "333": "Lighter", "334": "Grapefruit", "335": "Game board", "336": "Mop", "337": "Radish", "338": "Baozi", "339": "Target", "340": "French", "341": "Spring Rolls", "342": "Monkey", "343": "Rabbit", "344": "Pencil Case", "345": "Yak", "346": "Red Cabbage", "347": "Binoculars", "348": "Asparagus", "349": "Barbell", "350": "Scallop", "351": "Noddles", "352": "Comb", "353": "Dumpling", "354": "Oyster", "355": "Table Tennis paddle", "356": "Cosmetics Brush/Eyeliner Pencil", "357": "Chainsaw", "358": "Eraser", "359": "Lobster", "360": "Durian", "361": "Okra", "362": "Lipstick", "363": "Cosmetics Mirror", "364": "Curling", "365": "Table Tennis" }, "init_std": 0.02, "init_xavier_std": 1.0, "is_encoder_decoder": true, "label2id": { "Air Conditioner": 84, "Airplane": 115, "Ambulance": 249, "American Football": 207, "Antelope": 319, "Apple": 83, "Asparagus": 348, "Avocado": 289, "Awning": 75, "Backpack": 39, "Balloon": 91, "Banana": 113, "Baozi": 338, "Barbell": 349, "Barrel/bucket": 49, "Baseball": 166, "Baseball Bat": 138, "Baseball Glove": 114, "Basket": 53, "Basketball": 178, "Bathtub": 192, "Bear": 296, "Bed": 76, "Belt": 37, "Bench": 25, "Bicycle": 47, "Billiards": 190, "Binoculars": 347, "Blackboard/Whiteboard": 102, "Blender": 236, "Board Eraser": 298, "Boat": 22, "Book": 19, "Boots": 30, "Bottle": 9, "Bow Tie": 152, "Bowl/Basin": 27, "Bracelet": 15, "Bread": 63, "Briefcase": 195, "Broccoli": 142, "Broom": 270, "Brush": 257, "Bus": 56, "Butterfly": 322, "CD": 285, "Cabbage": 234, "Cabinet/shelf": 13, "Cake": 98, "Calculator": 317, "Camel": 308, "Camera": 64, "Candle": 72, "Candy": 188, "Canned": 65, "Car": 6, "Carpet": 61, "Carriage": 216, "Carrot": 153, "Cat": 140, "Cell Phone": 62, "Cello": 230, "Chainsaw": 357, "Chair": 3, "Cheese": 281, "Cherry": 283, "Chicken": 222, "Chips": 304, "Chopsticks": 163, "Cigar/Cigarette": 197, "Cleaning Products": 162, "Clock": 95, "Coconut": 299, "Coffee Machine": 214, "Coffee Table": 168, "Comb": 352, "Computer Box": 160, "Converter": 191, "Cookies": 175, "Corn": 260, "Cosmetics": 245, "Cosmetics Brush/Eyeliner Pencil": 356, "Cosmetics Mirror": 363, "Couch": 51, "Cow": 97, "Crab": 312, "Crane": 132, "Crosswalk Sign": 306, "Cucumber": 196, "Cue": 288, "Cup": 11, "Curling": 364, "Cutting/chopping Board": 167, "Cymbal": 68, "Deer": 241, "Desk": 10, "Dessert": 129, "Dinning Table": 99, "Dishwasher": 311, "Dog": 93, "Dolphin": 327, "Donkey": 324, "Donut": 151, "Drum": 54, "Duck": 137, "Dumbbell": 323, "Dumpling": 353, "Durian": 360, "Egg": 186, "Egg tart": 330, "Eggplant": 297, "Electric Drill": 328, "Elephant": 145, "Eraser": 358, "Extension Cord": 203, "Extractor": 202, "Fan": 111, "Faucet": 77, "Fire Extinguisher": 187, "Fire Hydrant": 177, "Fire Truck": 189, "Fishing Rod": 254, "Flag": 28, "Flask": 291, "Flower": 24, "Flute": 256, "Folder": 206, "Fork": 89, "Formula 1": 309, "French": 340, "French Fries": 231, "Frisbee": 220, "Game board": 335, "Garlic": 262, "Gas stove": 150, "Giraffe": 181, "Glasses": 8, "Globe": 303, "Gloves": 20, "Goldfish": 274, "Golf Ball": 248, "Golf Club": 194, "Goose": 242, "Grape": 180, "Grapefruit": 334, "Green Onion": 265, "Green Vegetables": 112, "Green beans": 218, "Guitar": 60, "Gun": 148, "Hair Dryer": 329, "Hamburger": 201, "Hamimelon": 290, "Hammer": 287, "Handbag/Satchel": 14, "Hanger": 101, "Hat": 5, "Head Phone": 126, "Heavy Truck": 200, "Helicopter": 264, "Helmet": 18, "High Heels": 58, "Hockey Stick": 86, "Horse": 79, "Hot dog": 235, "Hot-air balloon": 229, "Hoverboard": 313, "Hurdle": 253, "Ice cream": 228, "Induction Cooker": 269, "Jellyfish": 331, "Jug": 141, "Kettle": 210, "Key": 252, "Keyboard": 107, "Kite": 155, "Kiwi fruit": 275, "Knife": 85, "Ladder": 173, "Lamp": 7, "Lantern": 109, "Laptop": 74, "Leather Shoes": 23, "Lemon": 136, "Lettuce": 261, "Lifesaver": 69, "Lighter": 333, "Lion": 325, "Lipstick": 362, "Lobster": 359, "Luggage": 121, "Machinery Vehicle": 110, "Mango": 251, "Marker": 171, "Mask": 209, "Meat ball": 314, "Medal": 255, "Megaphone": 259, "Microphone": 32, "Microwave": 164, "Mirror": 80, "Monitor/TV": 38, "Monkey": 342, "Mop": 336, "Motorcycle": 59, "Mouse": 116, "Mushroom": 292, "Napkin": 103, "Necklace": 33, "Nightstand": 122, "Noddles": 351, "None": 0, "Notepaper": 282, "Nuts": 267, "Okra": 361, "Onion": 217, "Orange/Tangerine": 105, "Other Balls": 157, "Other Fish": 104, "Other Shoes": 4, "Oven": 135, "Oyster": 354, "Paddle": 87, "Paint Brush": 198, "Papaya": 318, "Parking meter": 250, "Parrot": 320, "Pasta": 286, "Peach": 237, "Pear": 199, "Pen/Pencil": 55, "Pencil Case": 344, "Penguin": 258, "Pepper": 159, "Person": 1, "Piano": 143, "Pickup Truck": 88, "Picture/Frame": 17, "Pie": 172, "Pig": 301, "Pigeon": 165, "Pillow": 29, "Pineapple": 247, "Pizza": 144, "Plate": 16, "Pliers": 284, "Plum": 272, "Poker Card": 277, "Pomegranate": 310, "Pot": 96, "Potato": 182, "Potted Plant": 26, "Power outlet": 81, "Printer": 223, "Projector": 219, "Pumpkin": 118, "Rabbit": 343, "Radiator": 176, "Radish": 337, "Recorder": 295, "Red Cabbage": 346, "Refrigerator": 134, "Remote": 133, "Rice": 238, "Rice Cooker": 315, "Rickshaw": 273, "Ring": 34, "Router/modem": 276, "SUV": 35, "Sailboat": 73, "Sandals": 52, "Sandwich": 266, "Sausage": 183, "Saxophone": 225, "Scale": 232, "Scallop": 350, "Scissors": 170, "Scooter": 130, "Screwdriver": 293, "Seal": 321, "Sheep": 100, "Ship": 212, "Shovel": 158, "Showerhead": 302, "Shrimp": 279, "Side Table": 169, "Sink": 82, "Skateboard": 146, "Skating and Skiing shoes": 149, "Skiboard": 120, "Slide": 215, "Slippers": 46, "Sneakers": 2, "Snowboard": 174, "Soap": 294, "Soccer": 119, "Speaker": 42, "Speed Limit Sign": 268, "Spoon": 94, "Sports Car": 127, "Spring Rolls": 341, "Stapler": 307, "Steak": 305, "Stool": 48, "Stop Sign": 128, "Storage box": 21, "Strawberry": 156, "Street Lights": 12, "Stroller": 131, "Stuffed Toy": 71, "Surfboard": 147, "Surveillance Camera": 139, "Sushi": 280, "Swan": 263, "Swing": 213, "Table Tennis": 365, "Table Tennis paddle": 355, "Tablet": 244, "Tape": 243, "Tape Measure/Ruler": 300, "Target": 339, "Tea pot": 123, "Telephone": 124, "Tennis": 211, "Tennis Racket": 205, "Tent": 78, "Tie": 44, "Tissue": 226, "Toaster": 278, "Toilet": 154, "Toilet Paper": 161, "Toiletry": 106, "Tomato": 108, "Tong": 204, "Toothbrush": 227, "Towel": 70, "Traffic Light": 41, "Traffic Sign": 90, "Traffic cone": 67, "Train": 117, "Trash bin Can": 45, "Treadmill": 332, "Tricycle": 184, "Tripod": 92, "Trolley": 125, "Trombone": 271, "Trophy": 233, "Truck": 66, "Trumpet": 246, "Tuba": 316, "Umbrella": 40, "Urinal": 326, "Van": 50, "Vase": 31, "Violin": 185, "Volleyball": 240, "Wallet/Purse": 239, "Washing Machine/Drying Machine": 221, "Watch": 43, "Watermelon": 224, "Wheelchair": 193, "Wild Bird": 57, "Wine Glass": 36, "Yak": 345, "Zebra": 179, "earphone": 208 }, "mask_loss_coefficient": 1, "max_position_embeddings": 2048, "model_type": "deta", "num_feature_levels": 5, "num_queries": 900, "position_embedding_type": "sine", "torch_dtype": "float32", "transformers_version": null, "two_stage": true, "two_stage_num_proposals": 300, "with_box_refine": true }