Real Time Semantic Segmentation On Cityscapes
评估指标
Frame (fps)
mIoU
评测结果
各个模型在此基准测试上的表现结果
比较表格
| 模型名称 | Frame (fps) | mIoU |
|---|---|---|
| rethinking-bisenet-for-real-time-semantic | 97.0(1080Ti) | 76.8% |
| template-based-automatic-search-of-compact | 19 | 67.7% |
| semantic-image-segmentation-with-deep | 0.25 | 63.1% |
| efficient-dense-modules-of-asymmetric | 108.7 (1080Ti) | 67.3 |
| deep-dual-resolution-networks-for-real-time | 101.6(2080Ti) | 77.4% |
| hardnet-a-low-memory-traffic-network | - | - |
| 模型 7 | 163.9 | 71.5% |
| icnet-for-real-time-semantic-segmentation-on | 30.3 | 70.6% |
| pp-liteseg-a-superior-real-time-semantic | 195.3(1080Ti) | 73.9% |
| pp-liteseg-a-superior-real-time-semantic | 102.6(1080Ti) | 77.5% |
| multi-path-segmentation-network | 59.2 | 74.8% |
| pidnet-a-real-time-semantic-segmentation | 31.1(3090) | 80.6% |
| esnet-an-efficient-symmetric-network-for-real | 63 | 70.7% |
| enet-a-deep-neural-network-architecture-for | 76.9 | 58.3% |
| the-lovasz-softmax-loss-a-tractable-surrogate | 76.9 | 63.1% |
| temporally-distributed-networks-for-fast | 47.6 (Titan X) | 74.9% |
| bisenet-v2-bilateral-network-with-guided | 156 | 72.6% |
| rethink-dilated-convolution-for-real-time | 30 | 78.3% |
| pp-liteseg-a-superior-real-time-semantic | 273.6(1080Ti) | 72.0% |
| hyperseg-patch-wise-hypernetwork-for-real | 36.9 | 75.8% |
| pp-liteseg-a-superior-real-time-semantic | 143.6(1080Ti) | 74.9% |
| segblocks-block-based-dynamic-resolution | 48.6 (1080Ti) | 73.8% |
| template-based-automatic-search-of-compact | 10 | 67.8% |
| rethinking-bisenet-for-real-time-semantic | 188.6 | 73.4% |
| conditional-random-fields-as-recurrent-neural-1 | 1.4 | 62.5% |
| rethinking-bisenet-for-real-time-semantic | 250.4(1080Ti) | 71.9% |
| lednet-a-lightweight-encoder-decoder-network | 71 | 70.6% |
| in-defense-of-pre-trained-imagenet | 39.9 | 75.5% |
| bisenet-v2-bilateral-network-with-guided | 47.3 | 75.3% |
| bisenet-bilateral-segmentation-network-for | 65.5 | 74.7% |
| full-resolution-residual-networks-for | 2.1 | 71.8% |
| rethinking-bisenet-for-real-time-semantic | 126.7 | 75.3% |
| pidnet-a-real-time-semantic-segmentation | 42.2(3090) | 79.8% |
| incorporating-luminance-depth-and-color | 18.4 (1080Ti) | 71.3% |
| bisenet-bilateral-segmentation-network-for | 105.8 | 68.4% |
| bisenet-bilateral-segmentation-network-for | 65.5 | 74.7% |
| swinmtl-a-shared-architecture-for | - | 76.41% |
| semantic-flow-for-fast-and-accurate-scene | 25.7(1080Ti) | 80.4% |
| pidnet-a-real-time-semantic-segmentation | 93.2(3090) | 78.6% |