Final rankings for the 2019 MicroNet Challenge. Within each task, we recognize entries that finished in the top 10% of our parameter storage and math operation metrics as “Highly Storage Efficient” and “Highly Compute Efficient” respectively. These distinctions are marked by superscript “S” and “C” for storage and computation respectively. Entries marked as “verified” were manually reviewed by the organizers for correctness. All entries achieved the specified accuracy threshold for the task unless otherwise specified.
| Entry Name | Affiliation | GitHub | MicroNet Score | Verified |
|---|---|---|---|---|
| RIAIRSC | NLPR & AIRIA, Institute of Automation, Chinese Academy of Sciences | link | 0.1295 | Yes |
| RockerSC | Beihang University, Chinese Academy of Sciences, Zhejiang University | link | 0.1887 | Yes |
| Rocker | Beihang University, Chinese Academy of Sciences, Zhejiang University | link | 0.2053 | Yes |
| QualcommAI-MixNet | Qualcomm AI Research | link | 0.2968 | Yes |
| dISTillers | IST Austria & NeuralMagic, Inc. | link | 0.3772 | Yes |
| QualcommAI-EfficinetNet | Qualcomm AI Research | link | 0.3789 | Yes |
| Expasoft | Expasoft | link | 0.3975 | |
| Texas-EIC | Rice University | link | 0.4678 | Yes |
| DQStarter | link | 0.5067* | ||
| QualcommAI-M0 | Qualcomm AI Research | link | 0.5477 | Yes |
| Alessandro Pappalardo | Xilinx Research Labs | link | 0.5558 | |
| DQStarter | link | 0.5789 | ||
| HHI-MAL | Fraunhofer Heinrich Hertz Institute | link | 0.5818 | |
| MB-PM Research | link | 0.6200 | ||
| Sisyphus | link | 0.6425 | Yes | |
| Expasoft | Expasoft | link | 0.6684 | |
| AutoPrune (A*STAR) | link | 0.8281 | ||
| Discovering Neural Wirings Team | AI2, University of Washington, XNOR.AI | link | 0.8460 | |
| Sisyphus | link | 1.0315 | Yes | |
| deep_learning_zs | link | 1.1957 |
| Entry Name | Affiliation | GitHub | MicroNet Score | Verified |
|---|---|---|---|---|
| RIAIRS | NLPR & AIRIA, Institute of Automation, Chinese Academy of Sciences | link | 0.0044 | Yes |
| KAIST AISC | KAIST | link | 0.0054 | Yes |
| KAIST AIC | KAIST | link | 0.0056 | Yes |
| OSI AIC | KAIST | link | 0.0058 | Yes |
| MSUNet-V3S | Michigan State University | link | 0.0108 | Yes |
| MSUNet-V2 | Michigan State University | link | 0.0125 | Yes |
| MSUNet-V1 | Michigan State University | link | 0.0171 | Yes |
| Woody | Northestern University, Indiana Univeristy, MIT-IBM Watson AI Lab | link | 0.0188 | Yes |
| Woody | Northestern University, Indiana Univeristy, MIT-IBM Watson AI Lab | link | 0.0193 | Yes |
| HHI-MAL | Fraunhofer Heinrich Hertz Institute | link | 0.0242 | Yes |
| Woody | Northestern University, Indiana Univeristy, MIT-IBM Watson AI Lab | link | 0.0266 | Yes |
| MB-PM Research | link | 0.0384 | ||
| WaveComp | Wave Computing | link | 0.0456 | |
| WaveComp | Wave Computing | link | 0.0464 | |
| WaveComp | Wave Computing | link | 0.0467 | |
| Sloth | link | 0.0618 | ||
| MB-PM Research | link | 0.0653 | ||
| QualcommAI-nanoWRN | Qualcomm AI Research | link | 0.0721 | |
| BrAIn | IMT Atlantique | link | 0.0780 | |
| UCI-M | University of California, Irvine | link | 0.1114 | |
| AutoPrune (A*STAR) | link | 0.1118 | ||
| UCI-M | University of California, Irvine | link | 0.1254 | |
| Frenzy | ARM ML Research | link | 0.1368 | |
| ProxylessNAS-TTQ | link | 0.1433 | ||
| micro-machines | link | 0.1659 | ||
| PunyNet | IIT Roorkee | link | 0.1906* | |
| ID56 | link | 0.2056 | ||
| Soft Binary Mask | link | 0.3847 | ||
| IamShinichiYamamoto | link | 0.4635 | ||
| Sloth | link | 0.7635 |
| Entry Name | Affiliation | GitHub | MicroNet Score | Verified |
|---|---|---|---|---|
| MIT-HAN-LabC | MIT | link | 0.0475 | Yes |
| MIT-HAN-LabS | MIT | link | 0.0482 | Yes |
| MIT-HAN-Lab | MIT | link | 0.0485 | Yes |
| Clova AI | Clova AI/Kyoto University | link | 0.1657 | Yes |
| JNLP | JAIST/ISM | link | 0.8232 | Yes |
*Did not meet required accuracy target for the task. PunyNet CIFAR-100 entry achieved 53.39% top-1 accuracy. DQStarter ImageNet entry achieved 74.962% top-1.