## 图像长宽统计
大部分都是 `800*800` 的 尺寸，少部分会非常大
## 统计小目标的比例
```
+------------------------+---------------------+
| class | small object ratio |
+------------------------+---------------------+
| 冰墩墩 | 0.35398230088495575 |
| Sanyo/三洋 | 1.0 |
| Eifini/伊芙丽 | 1.0 |
| PSALTER/诗篇 | 0.9636363636363636 |
| Beaster | 0.84375 |
| ON/昂跑 | 0.9878048780487805 |
| BYREDO/柏芮朵 | 0.9568965517241379 |
| Ubras | 0.98 |
| Eternelle | 0.6326530612244898 |
| PERFECT DIARY/完美日记 | 0.8918918918918919 |
| 花西子 | 0.9897959183673469 |
| Clarins/娇韵诗 | 0.9734513274336283 |
| L'occitane/欧舒丹 | 0.9716312056737588 |
| Versace/范思哲 | 0.8235294117647058 |
| Mizuno/美津浓 | 0.7520661157024794 |
| Lining/李宁 | 0.95 |
| DOUBLE STAR/双星 | 0.5416666666666666 |
| YONEX/尤尼克斯 | 0.8475609756097561 |
| Tory Burch/汤丽柏琦 | 0.9105691056910569 |
| Gucci/古驰 | 0.9432624113475178 |
| Louis Vuitton/路易威登 | 0.9702970297029703 |
| CARTELO/卡帝乐鳄鱼 | 0.7894736842105263 |
| JORDAN | 0.828125 |
| KENZO | 0.8148148148148148 |
| UNDEFEATED | 0.8936170212765957 |
| BOY LONDON | 0.6715328467153284 |
| TREYO/雀友 | 0.9081632653061225 |
| carhartt | 0.9514563106796117 |
| 洁柔 | 0.9771241830065359 |
| Blancpain/宝珀 | 1.0 |
| GXG | 1.0 |
| 乐町 | 1.0 |
| Diadora/迪亚多纳 | 0.38271604938271603 |
| TUCANO/啄木鸟 | 0.6031746031746031 |
| Loewe | 0.9120879120879121 |
| Granite Gear | 0.9813084112149533 |
| DESCENTE/迪桑特 | 1.0 |
| OSPREY | 0.8968253968253969 |
| Swatch/斯沃琪 | 0.9466666666666667 |
| erke/鸿星尔克 | 0.8674698795180723 |
| Massimo Dutti | 0.9807692307692307 |
| PINKO | 0.8390804597701149 |
| PALLADIUM | 0.9441624365482234 |
| origins/悦木之源 | 0.9767441860465116 |
| Trendiano | 1.0 |
| 音儿 | 1.0 |
| Monster Guardians | 0.9891304347826086 |
| 敷尔佳 | 0.8620689655172413 |
| IPSA/茵芙莎 | 0.9777777777777777 |
| Schwarzkopf/施华蔻 | 0.954954954954955 |
| all | 0.8899438093392753 |
+------------------------+---------------------+
```
小目标(area<%3) 占据了大部分
## bbox长宽比
LogDet-master.zip （19个子文件） 
LogDet-master ContestStage1 train configs 
swinb_3x_bigsize_anchor_bs2x8.py 10KB data 
class_names.txt 699B instances_train2017.json 1.02MB instances_val2017.json 264KB eda readme.md 3KB logdet-eda.ipynb 4.22MB LogDetMini train configs LogDet r101_12x.py 9KB r101.py 9KB swinb_3x_800-1400_anchor_bs2x8.py 10KB r101_3x_flip.py 9KB r101_3x_flip_softnms.py 9KB r101_3x.py 9KB readme.md 218B logdet_mini.py 23KB single_level_roi_extractor.py 5KB readme.md 903B eda readme.md 3KB eda.ipynb 34KB README.md 2KB资源评论
