Semantic Segmentation on Pascal Context, Pascal VOC, Pascal Stuff, ADE20K, and Cityscapes Average (val)
50.01mIoURADSeg+
Evaluation Results
| Method | Links | |
|---|---|---|
| RADSeg+Model Scale=Base, Latency (CTX)=0.261, Latency (VOC)=0.156, Latency (Stuff)=0.286, Latency (ADE)=0.360, Latency (City)=0.709, Average Latency (s)=0.354, Parameters (M)=119, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 50.01 | |
| RADSeg+Model Scale=Huge, Latency (CTX)=1.172, Latency (VOC)=0.734, Latency (Stuff)=1.074, Latency (ADE)=1.341, Latency (City)=2.128, Average Latency (s)=1.290, Parameters (M)=677, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 49.38 | |
| RADSegModel Scale=Base, Latency (CTX)=0.109, Latency (VOC)=0.036, Latency (Stuff)=0.088, Latency (ADE)=0.112, Latency (City)=0.228, Average Latency (s)=0.115, Parameters (M)=106, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 48 | |
| TextRegionModel Scale=Huge, Latency (CTX)=0.732, Latency (VOC)=0.670, Latency (Stuff)=0.711, Latency (ADE)=0.744, Latency (City)=2.476, Average Latency (s)=1.067, Parameters (M)=857, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 47.5 | |
| RADSegModel Scale=Huge, Latency (CTX)=0.617, Latency (VOC)=0.202, Latency (Stuff)=0.457, Latency (ADE)=0.594, Latency (City)=1.363, Average Latency (s)=0.647, Parameters (M)=664, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 46.92 | |
| TridentModel Scale=Huge, Latency (CTX)=1.426, Latency (VOC)=0.900, Latency (Stuff)=1.123, Latency (ADE)=1.444, Latency (City)=2.365, Average Latency (s)=1.451, Parameters (M)=1349, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 46.9 | |
| TridentModel Scale=Base, Latency (CTX)=0.353, Latency (VOC)=0.229, Latency (Stuff)=0.348, Latency (ADE)=0.414, Latency (City)=0.923, Average Latency (s)=0.454, Parameters (M)=264, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 43.79 | |
| TextRegionModel Scale=Base, Latency (CTX)=0.410, Latency (VOC)=0.403, Latency (Stuff)=0.438, Latency (ADE)=0.428, Latency (City)=1.729, Average Latency (s)=0.682, Parameters (M)=167, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 43.33 | |
| ProxyCLIPModel Scale=Huge, Latency (CTX)=1.297, Latency (VOC)=0.494, Latency (Stuff)=0.451, Latency (ADE)=1.28, Latency (City)=3.168, Average Latency (s)=1.338, Parameters (M)=717, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 43.21 | |
| Talk2DinoModel Scale=Base, Latency (CTX)=0.150, Latency (VOC)=0.043, Latency (Stuff)=0.155, Latency (ADE)=0.188, Latency (City)=0.289, Average Latency (s)=0.165, Parameters (M)=86, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 43 | |
| RayFrontsModel Scale=Base, Latency (CTX)=0.106, Latency (VOC)=0.035, Latency (Stuff)=0.087, Latency (ADE)=0.109, Latency (City)=0.218, Average Latency (s)=0.111, Parameters (M)=103, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 42.14 | |
| SC-CLIPModel Scale=Base, Latency (CTX)=0.148, Latency (VOC)=0.059, Latency (Stuff)=0.111, Latency (ADE)=0.174, Latency (City)=0.243, Average Latency (s)=0.147, Parameters (M)=86, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 41.86 | |
| ProxyCLIPModel Scale=Base, Latency (CTX)=0.689, Latency (VOC)=0.262, Latency (Stuff)=0.247, Latency (ADE)=0.693, Latency (City)=1.446, Average Latency (s)=0.667, Parameters (M)=171, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 40.87 | |
| ResCLIPModel Scale=Base, Latency (CTX)=0.252, Latency (VOC)=0.152, Latency (Stuff)=0.266, Latency (ADE)=0.325, Latency (City)=0.311, Average Latency (s)=0.261, Parameters (M)=86, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 40.33 | |
| NACLIPModel Scale=Base, Latency (CTX)=0.091, Latency (VOC)=0.025, Latency (Stuff)=0.074, Latency (ADE)=0.111, Latency (City)=0.144, Average Latency (s)=0.089, Parameters (M)=86, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 38.48 | |
| RayFrontsModel Scale=Huge, Latency (CTX)=0.619, Latency (VOC)=0.204, Latency (Stuff)=0.460, Latency (ADE)=0.598, Latency (City)=1.362, Average Latency (s)=0.649, Parameters (M)=661, Resolution=Mid, Precision=FP32, Hardware=Tesla V1002025.11 | 36.4 |