表中模型相关指标均为在ImageNet数据集上使用PaddlePaddle Python预测接口测试得到(测试GPU型号为Nvidia Tesla P40),预测速度为每张图片预测用时(不包括预处理和后处理),表中符号
-表示相关指标暂未测试。
| 模型 | 模型大小 | 预测速度(毫秒) | Top1准确率(%) | Top5准确率(%) |
|---|---|---|---|---|
| ResNet18 | 46.2MB | 3.72882 | 71.0 | 89.9 |
| ResNet34 | 87.9MB | 5.50876 | 74.6 | 92.1 |
| ResNet50 | 103.4MB | 7.76659 | 76.5 | 93.0 |
| ResNet101 | 180.4MB | 13.80876 | 77.6 | 93.6 |
| ResNet50_vd | 103.5MB | 8.20476 | 79.1 | 94.4 |
| ResNet101_vd | 180.5MB | 14.24643 | 80.2 | 95.0 |
| ResNet50_vd_ssld | 103.5MB | 7.79264 | 82.4 | 96.1 |
| ResNet101_vd_ssld | 180.5MB | 13.34580 | 83.7 | 96.7 |
| DarkNet53 | 167.4MB | 8.82047 | 78.0 | 94.1 |
| MobileNetV1 | 17.4MB | 3.42838 | 71.0 | 89.7 |
| MobileNetV2 | 15.0MB | 5.92667 | 72.2 | 90.7 |
| MobileNetV3_large | 22.8MB | 8.31428 | 75.3 | 93.2 |
| MobileNetV3_small | 12.5MB | 7.30689 | 68.2 | 88.1 |
| MobileNetV3_large_ssld | 22.8MB | 8.06651 | 79.0 | 94.5 |
| MobileNetV3_small_ssld | 12.5MB | 7.08837 | 71.3 | 90.1 |
| Xception41 | 109.2MB | 8.15611 | 79.6 | 94.4 |
| Xception65 | 161.6MB | 13.87017 | 80.3 | 94.5 |
| DenseNet121 | 33.1MB | 17.09874 | 75.7 | 92.6 |
| DenseNet161 | 118.0MB | 22.79690 | 78.6 | 94.1 |
| DenseNet201 | 84.1MB | 25.26089 | 77.6 | 93.7 |
| ShuffleNetV2 | 10.2MB | 15.40138 | 68.8 | 88.5 |
| HRNet_W18 | 21.29MB | 45.25514 | 76.9 | 93.4 |
表中模型相关指标均为在MSCOCO数据集上使用PaddlePaddle Python预测接口测试得到(测试GPU型号为Nvidia Tesla P40测试得到,表中符号
-表示相关指标暂未测试。
| 模型 | 模型大小 | 预测时间(毫秒) | BoxAP(%) |
|---|---|---|---|
| FasterRCNN-ResNet50 | 136.0MB | 316.912 | 35.2 |
| FasterRCNN-ResNet50_vd | 136.1MB | 302.495 | 36.4 |
| FasterRCNN-ResNet101 | 212.5MB | 339.153 | 38.3 |
| FasterRCNN-ResNet50-FPN | 167.7MB | 44.897 | 37.2 |
| FasterRCNN-ResNet50_vd-FPN | 167.8MB | 73.219 | 38.9 |
| FasterRCNN-ResNet101-FPN | 244.2MB | 93.236 | 38.7 |
| FasterRCNN-ResNet101_vd-FPN | 244.3MB | 96.424 | 40.5 |
| FasterRCNN-HRNet_W18-FPN | 115.5MB | 81.592 | 36 |
| YOLOv3-DarkNet53 | 249.2MB | 320.738 | 38.9 |
| YOLOv3-MobileNetV1 | 99.2MB | 349.461 | 29.3 |
| YOLOv3-MobileNetV3_large | 100.7MB | 480.075 | 31.6 |
| YOLOv3-ResNet34 | 170.3MB | 417.680 | 36.2 |
表中模型相关指标均为在MSCOCO数据集上测试得到。
| 模型 | 模型大小 | 预测时间(毫秒) | mIoU(%) |
|---|---|---|---|
| DeepLabv3+-MobileNetV2_x1.0 | - | - | - |
| DeepLabv3+-Xception41 | - | - | - |
| DeepLabv3+-Xception65 | - | - | - |
| UNet | - | - | - |
| HRNet_w18 | - | - | - |