Upload MaskRCNNForObjectDetection
Browse files- config.json +16 -81
config.json
CHANGED
@@ -19,40 +19,9 @@
|
|
19 |
"MaskRCNNForObjectDetection"
|
20 |
],
|
21 |
"backbone_config": {
|
22 |
-
"_name_or_path": "",
|
23 |
-
"add_cross_attention": false,
|
24 |
"architectures": [
|
25 |
"ConvNextForImageClassification"
|
26 |
],
|
27 |
-
"bad_words_ids": null,
|
28 |
-
"begin_suppress_tokens": null,
|
29 |
-
"bos_token_id": null,
|
30 |
-
"chunk_size_feed_forward": 0,
|
31 |
-
"cross_attention_hidden_size": null,
|
32 |
-
"decoder_start_token_id": null,
|
33 |
-
"depths": [
|
34 |
-
3,
|
35 |
-
3,
|
36 |
-
9,
|
37 |
-
3
|
38 |
-
],
|
39 |
-
"diversity_penalty": 0.0,
|
40 |
-
"do_sample": false,
|
41 |
-
"drop_path_rate": 0.0,
|
42 |
-
"early_stopping": false,
|
43 |
-
"encoder_no_repeat_ngram_size": 0,
|
44 |
-
"eos_token_id": null,
|
45 |
-
"exponential_decay_length_penalty": null,
|
46 |
-
"finetuning_task": null,
|
47 |
-
"forced_bos_token_id": null,
|
48 |
-
"forced_eos_token_id": null,
|
49 |
-
"hidden_act": "gelu",
|
50 |
-
"hidden_sizes": [
|
51 |
-
96,
|
52 |
-
192,
|
53 |
-
384,
|
54 |
-
768
|
55 |
-
],
|
56 |
"id2label": {
|
57 |
"0": "tench, Tinca tinca",
|
58 |
"1": "goldfish, Carassius auratus",
|
@@ -1055,10 +1024,6 @@
|
|
1055 |
"998": "ear, spike, capitulum",
|
1056 |
"999": "toilet tissue, toilet paper, bathroom tissue"
|
1057 |
},
|
1058 |
-
"image_size": 224,
|
1059 |
-
"initializer_range": 0.02,
|
1060 |
-
"is_decoder": false,
|
1061 |
-
"is_encoder_decoder": false,
|
1062 |
"label2id": {
|
1063 |
"Afghan hound, Afghan": 160,
|
1064 |
"African chameleon, Chamaeleo chamaeleon": 47,
|
@@ -2060,58 +2025,20 @@
|
|
2060 |
"zebra": 340,
|
2061 |
"zucchini, courgette": 939
|
2062 |
},
|
2063 |
-
"layer_norm_eps": 1e-12,
|
2064 |
-
"layer_scale_init_value": 1e-06,
|
2065 |
-
"length_penalty": 1.0,
|
2066 |
-
"max_length": 20,
|
2067 |
-
"min_length": 0,
|
2068 |
"model_type": "convnext",
|
2069 |
-
"no_repeat_ngram_size": 0,
|
2070 |
-
"num_beam_groups": 1,
|
2071 |
-
"num_beams": 1,
|
2072 |
-
"num_channels": 3,
|
2073 |
-
"num_return_sequences": 1,
|
2074 |
-
"num_stages": 4,
|
2075 |
"out_features": [
|
2076 |
"stage1",
|
2077 |
"stage2",
|
2078 |
"stage3",
|
2079 |
"stage4"
|
2080 |
],
|
2081 |
-
"
|
2082 |
-
|
2083 |
-
|
2084 |
-
|
2085 |
-
|
2086 |
-
"prefix": null,
|
2087 |
-
"problem_type": null,
|
2088 |
-
"pruned_heads": {},
|
2089 |
-
"remove_invalid_values": false,
|
2090 |
-
"repetition_penalty": 1.0,
|
2091 |
-
"return_dict": true,
|
2092 |
-
"return_dict_in_generate": false,
|
2093 |
-
"sep_token_id": null,
|
2094 |
-
"stage_names": [
|
2095 |
-
"stem",
|
2096 |
-
"stage1",
|
2097 |
-
"stage2",
|
2098 |
-
"stage3",
|
2099 |
-
"stage4"
|
2100 |
],
|
2101 |
-
"
|
2102 |
-
"task_specific_params": null,
|
2103 |
-
"temperature": 1.0,
|
2104 |
-
"tf_legacy_loss": false,
|
2105 |
-
"tie_encoder_decoder": false,
|
2106 |
-
"tie_word_embeddings": true,
|
2107 |
-
"tokenizer_class": null,
|
2108 |
-
"top_k": 50,
|
2109 |
-
"top_p": 1.0,
|
2110 |
-
"torch_dtype": "float32",
|
2111 |
-
"torchscript": false,
|
2112 |
-
"transformers_version": "4.28.0.dev0",
|
2113 |
-
"typical_p": 1.0,
|
2114 |
-
"use_bfloat16": false
|
2115 |
},
|
2116 |
"bbox_head_bbox_coder_target_means": [
|
2117 |
0.0,
|
@@ -2125,7 +2052,12 @@
|
|
2125 |
0.2,
|
2126 |
0.2
|
2127 |
],
|
|
|
2128 |
"bbox_head_in_channels": 256,
|
|
|
|
|
|
|
|
|
2129 |
"bbox_roi_extractor_featmap_strides": [
|
2130 |
4,
|
2131 |
8,
|
@@ -2305,6 +2237,9 @@
|
|
2305 |
"wine glass": 40,
|
2306 |
"zebra": 22
|
2307 |
},
|
|
|
|
|
|
|
2308 |
"mask_roi_extractor_featmap_strides": [
|
2309 |
4,
|
2310 |
8,
|
@@ -2324,8 +2259,8 @@
|
|
2324 |
"rcnn_assigner_neg_iou_thr": 0.5,
|
2325 |
"rcnn_assigner_pos_iou_thr": 0.5,
|
2326 |
"rcnn_sampler_add_gt_as_proposals": true,
|
2327 |
-
"rcnn_sampler_neg_pos_ub": -1,
|
2328 |
"rcnn_sampler_num": 512,
|
|
|
2329 |
"rcnn_sampler_pos_fraction": 0.25,
|
2330 |
"rcnn_test_cfg": {
|
2331 |
"mask_thr_binary": 0.5,
|
@@ -2379,8 +2314,8 @@
|
|
2379 |
"nms_pre": 2000
|
2380 |
},
|
2381 |
"rpn_sampler_add_gt_as_proposals": false,
|
2382 |
-
"rpn_sampler_neg_pos_ub": -1,
|
2383 |
"rpn_sampler_num": 256,
|
|
|
2384 |
"rpn_sampler_pos_fraction": 0.5,
|
2385 |
"rpn_test_cfg": {
|
2386 |
"max_per_img": 1000,
|
|
|
19 |
"MaskRCNNForObjectDetection"
|
20 |
],
|
21 |
"backbone_config": {
|
|
|
|
|
22 |
"architectures": [
|
23 |
"ConvNextForImageClassification"
|
24 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
"id2label": {
|
26 |
"0": "tench, Tinca tinca",
|
27 |
"1": "goldfish, Carassius auratus",
|
|
|
1024 |
"998": "ear, spike, capitulum",
|
1025 |
"999": "toilet tissue, toilet paper, bathroom tissue"
|
1026 |
},
|
|
|
|
|
|
|
|
|
1027 |
"label2id": {
|
1028 |
"Afghan hound, Afghan": 160,
|
1029 |
"African chameleon, Chamaeleo chamaeleon": 47,
|
|
|
2025 |
"zebra": 340,
|
2026 |
"zucchini, courgette": 939
|
2027 |
},
|
|
|
|
|
|
|
|
|
|
|
2028 |
"model_type": "convnext",
|
|
|
|
|
|
|
|
|
|
|
|
|
2029 |
"out_features": [
|
2030 |
"stage1",
|
2031 |
"stage2",
|
2032 |
"stage3",
|
2033 |
"stage4"
|
2034 |
],
|
2035 |
+
"out_indices": [
|
2036 |
+
1,
|
2037 |
+
2,
|
2038 |
+
3,
|
2039 |
+
4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2040 |
],
|
2041 |
+
"torch_dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2042 |
},
|
2043 |
"bbox_head_bbox_coder_target_means": [
|
2044 |
0.0,
|
|
|
2052 |
0.2,
|
2053 |
0.2
|
2054 |
],
|
2055 |
+
"bbox_head_fc_out_channels": 1024,
|
2056 |
"bbox_head_in_channels": 256,
|
2057 |
+
"bbox_head_num_shared_fcs": 2,
|
2058 |
+
"bbox_head_reg_class_agnostic": false,
|
2059 |
+
"bbox_head_reg_decoded_bbox": false,
|
2060 |
+
"bbox_head_roi_feat_size": 7,
|
2061 |
"bbox_roi_extractor_featmap_strides": [
|
2062 |
4,
|
2063 |
8,
|
|
|
2237 |
"wine glass": 40,
|
2238 |
"zebra": 22
|
2239 |
},
|
2240 |
+
"mask_head_conv_out_channels": 256,
|
2241 |
+
"mask_head_in_channels": 256,
|
2242 |
+
"mask_head_num_convs": 4,
|
2243 |
"mask_roi_extractor_featmap_strides": [
|
2244 |
4,
|
2245 |
8,
|
|
|
2259 |
"rcnn_assigner_neg_iou_thr": 0.5,
|
2260 |
"rcnn_assigner_pos_iou_thr": 0.5,
|
2261 |
"rcnn_sampler_add_gt_as_proposals": true,
|
|
|
2262 |
"rcnn_sampler_num": 512,
|
2263 |
+
"rcnn_sampler_num_samples_upper_bound": -1,
|
2264 |
"rcnn_sampler_pos_fraction": 0.25,
|
2265 |
"rcnn_test_cfg": {
|
2266 |
"mask_thr_binary": 0.5,
|
|
|
2314 |
"nms_pre": 2000
|
2315 |
},
|
2316 |
"rpn_sampler_add_gt_as_proposals": false,
|
|
|
2317 |
"rpn_sampler_num": 256,
|
2318 |
+
"rpn_sampler_num_samples_upper_bound": -1,
|
2319 |
"rpn_sampler_pos_fraction": 0.5,
|
2320 |
"rpn_test_cfg": {
|
2321 |
"max_per_img": 1000,
|