Upload config.json
Browse files- config.json +219 -0
config.json
ADDED
|
@@ -0,0 +1,219 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"norm_cfg": {
|
| 3 |
+
"type": "SyncBN",
|
| 4 |
+
"requires_grad": true
|
| 5 |
+
},
|
| 6 |
+
"model": {
|
| 7 |
+
"type": "EncoderDecoder",
|
| 8 |
+
"pretrained": "pretrained/swin_small_patch4_window7_224.pth",
|
| 9 |
+
"backbone": {
|
| 10 |
+
"type": "SwinTransformer",
|
| 11 |
+
"embed_dim": 96,
|
| 12 |
+
"depths": [2, 2, 18, 2],
|
| 13 |
+
"num_heads": [3, 6, 12, 24],
|
| 14 |
+
"window_size": 7,
|
| 15 |
+
"mlp_ratio": 4.0,
|
| 16 |
+
"qkv_bias": true,
|
| 17 |
+
"qk_scale": null,
|
| 18 |
+
"drop_rate": 0.0,
|
| 19 |
+
"attn_drop_rate": 0.0,
|
| 20 |
+
"drop_path_rate": 0.3,
|
| 21 |
+
"ape": false,
|
| 22 |
+
"patch_norm": true,
|
| 23 |
+
"out_indices": [0, 1, 2, 3],
|
| 24 |
+
"use_checkpoint": false
|
| 25 |
+
},
|
| 26 |
+
"decode_head": {
|
| 27 |
+
"type": "UPerHead",
|
| 28 |
+
"in_channels": [96, 192, 384, 768],
|
| 29 |
+
"in_index": [0, 1, 2, 3],
|
| 30 |
+
"pool_scales": [1, 2, 3, 6],
|
| 31 |
+
"channels": 512,
|
| 32 |
+
"dropout_ratio": 0.1,
|
| 33 |
+
"num_classes": 104,
|
| 34 |
+
"norm_cfg": {
|
| 35 |
+
"type": "SyncBN",
|
| 36 |
+
"requires_grad": true
|
| 37 |
+
},
|
| 38 |
+
"align_corners": false,
|
| 39 |
+
"loss_decode": {
|
| 40 |
+
"type": "CrossEntropyLoss",
|
| 41 |
+
"use_sigmoid": false,
|
| 42 |
+
"loss_weight": 1.0
|
| 43 |
+
}
|
| 44 |
+
},
|
| 45 |
+
"auxiliary_head": {
|
| 46 |
+
"type": "FCNHead",
|
| 47 |
+
"in_channels": 384,
|
| 48 |
+
"in_index": 2,
|
| 49 |
+
"channels": 256,
|
| 50 |
+
"num_convs": 1,
|
| 51 |
+
"concat_input": false,
|
| 52 |
+
"dropout_ratio": 0.1,
|
| 53 |
+
"num_classes": 104,
|
| 54 |
+
"norm_cfg": {
|
| 55 |
+
"type": "SyncBN",
|
| 56 |
+
"requires_grad": true
|
| 57 |
+
},
|
| 58 |
+
"align_corners": false,
|
| 59 |
+
"loss_decode": {
|
| 60 |
+
"type": "CrossEntropyLoss",
|
| 61 |
+
"use_sigmoid": false,
|
| 62 |
+
"loss_weight": 0.4
|
| 63 |
+
}
|
| 64 |
+
},
|
| 65 |
+
"train_cfg": {},
|
| 66 |
+
"test_cfg": {
|
| 67 |
+
"mode": "whole"
|
| 68 |
+
}
|
| 69 |
+
},
|
| 70 |
+
"dataset_type": "CustomDataset",
|
| 71 |
+
"data_root": "./data/FoodSeg103/Images/",
|
| 72 |
+
"img_norm_cfg": {
|
| 73 |
+
"mean": [123.675, 116.28, 103.53],
|
| 74 |
+
"std": [58.395, 57.12, 57.375],
|
| 75 |
+
"to_rgb": true
|
| 76 |
+
},
|
| 77 |
+
"crop_size": [512, 1024],
|
| 78 |
+
"train_pipeline": [
|
| 79 |
+
{
|
| 80 |
+
"type": "LoadImageFromFile"
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"type": "LoadAnnotations"
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"type": "Resize",
|
| 87 |
+
"img_scale": [2048, 1024],
|
| 88 |
+
"ratio_range": [0.5, 2.0]
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"type": "RandomCrop",
|
| 92 |
+
"crop_size": [512, 1024],
|
| 93 |
+
"cat_max_ratio": 0.75
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"type": "RandomFlip",
|
| 97 |
+
"prob": 0.5
|
| 98 |
+
},
|
| 99 |
+
{
|
| 100 |
+
"type": "PhotoMetricDistortion"
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"type": "Normalize",
|
| 104 |
+
"mean": [123.675, 116.28, 103.53],
|
| 105 |
+
"std": [58.395, 57.12, 57.375],
|
| 106 |
+
"to_rgb": true
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"type": "Pad",
|
| 110 |
+
"size": [512, 1024],
|
| 111 |
+
"pad_val": 0,
|
| 112 |
+
"seg_pad_val": 255
|
| 113 |
+
},
|
| 114 |
+
{
|
| 115 |
+
"type": "DefaultFormatBundle"
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"type": "Collect",
|
| 119 |
+
"keys": ["img", "gt_semantic_seg"]
|
| 120 |
+
}
|
| 121 |
+
],
|
| 122 |
+
"test_pipeline": [
|
| 123 |
+
{
|
| 124 |
+
"type": "LoadImageFromFile"
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"type": "MultiScaleFlipAug",
|
| 128 |
+
"img_scale": [2048, 1024],
|
| 129 |
+
"flip": false,
|
| 130 |
+
"transforms": [
|
| 131 |
+
{
|
| 132 |
+
"type": "Resize",
|
| 133 |
+
"keep_ratio": true
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"type": "RandomFlip"
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"type": "Normalize",
|
| 140 |
+
"mean": [123.675, 116.28, 103.53],
|
| 141 |
+
"std": [58.395, 57.12, 57.375],
|
| 142 |
+
"to_rgb": true
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"type": "ImageToTensor",
|
| 146 |
+
"keys": ["img"]
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"type": "Collect",
|
| 150 |
+
"keys": ["img"]
|
| 151 |
+
}
|
| 152 |
+
]
|
| 153 |
+
}
|
| 154 |
+
],
|
| 155 |
+
"data": {
|
| 156 |
+
"samples_per_gpu": 2,
|
| 157 |
+
"workers_per_gpu": 2,
|
| 158 |
+
"train": {
|
| 159 |
+
"type": "CustomDataset",
|
| 160 |
+
"data_root": "./data/FoodSeg103/Images/",
|
| 161 |
+
"img_dir": "img_dir/train",
|
| 162 |
+
"ann_dir": "ann_dir/train",
|
| 163 |
+
"pipeline": [
|
| 164 |
+
{
|
| 165 |
+
"type": "LoadImageFromFile"
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"type": "LoadAnnotations"
|
| 169 |
+
},
|
| 170 |
+
{
|
| 171 |
+
"type": "Resize",
|
| 172 |
+
"img_scale": [2048, 1024],
|
| 173 |
+
"ratio_range": [0.5, 2.0]
|
| 174 |
+
},
|
| 175 |
+
{
|
| 176 |
+
"type": "RandomCrop",
|
| 177 |
+
"crop_size": [512, 1024],
|
| 178 |
+
"cat_max_ratio": 0.75
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"type": "RandomFlip",
|
| 182 |
+
"prob": 0.5
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"type": "PhotoMetricDistortion"
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"type": "Normalize",
|
| 189 |
+
"mean": [123.675, 116.28, 103.53],
|
| 190 |
+
"std": [58.395, 57.12, 57.375],
|
| 191 |
+
"to_rgb": true
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"type": "Pad",
|
| 195 |
+
"size": [512, 1024],
|
| 196 |
+
"pad_val": 0,
|
| 197 |
+
"seg_pad_val": 255
|
| 198 |
+
},
|
| 199 |
+
{
|
| 200 |
+
"type": "DefaultFormatBundle"
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"type": "Collect",
|
| 204 |
+
"keys": ["img", "gt_semantic_seg"]
|
| 205 |
+
}
|
| 206 |
+
]
|
| 207 |
+
},
|
| 208 |
+
"val": {
|
| 209 |
+
"type": "CustomDataset",
|
| 210 |
+
"data_root": "./data/FoodSeg103/Images/",
|
| 211 |
+
"img_dir": "img_dir/test",
|
| 212 |
+
"ann_dir": "ann_dir/test",
|
| 213 |
+
"pipeline": [
|
| 214 |
+
{
|
| 215 |
+
"type": "LoadImageFromFile"
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"type": "MultiScaleFlipAug",
|
| 219 |
+
"img_scale": [2048, 1024],
|