| RTMO-s (body7) — acaua mirror (pure-PyTorch port) |
| ==================================================== |
| |
| This product includes: |
| |
| 1. PORTED SOURCE CODE: a pure-PyTorch port of the RTMO architecture |
| (located in the acaua repository at src/acaua/adapters/rtmo/) is a |
| derivative work of: |
| |
| - OpenMMLab's mmpose implementation |
| https://github.com/open-mmlab/mmpose @ commit |
| 759b39c13fea6ba094afc1fa932f51dc1b11cbf9 — Apache-2.0 |
| Files derived from: |
| mmpose/models/backbones/csp_darknet.py |
| mmpose/models/necks/hybrid_encoder.py |
| mmpose/models/heads/hybrid_heads/rtmo_head.py |
| mmpose/models/heads/hybrid_heads/yoloxpose_head.py |
| mmpose/models/utils/csp_layer.py |
| mmpose/models/utils/reparam_layers.py |
| mmpose/models/utils/transformer.py |
| mmpose/evaluation/functional/nms.py |
| - Plus mmcv primitives (ConvModule, DepthwiseSeparableConvModule, |
| FFN, MultiheadAttention, Scale) vendored as pure-PyTorch |
| equivalents. |
| |
| Paper: Peng Lu, Tao Jiang, Yining Li, Xiangtai Li, Kai Chen, |
| Wenming Yang, "RTMO: Towards High-Performance One-Stage Real-Time |
| Multi-Person Pose Estimation", CVPR 2024 (arXiv:2312.07526). |
|
|
| 2. CONVERTED WEIGHTS: the model.safetensors file in this mirror is a |
| key-remapped conversion of the upstream pretrained checkpoint: |
| |
| - upstream URL: https://download.openmmlab.com/mmpose/v1/projects/rtmo/rtmo-s_8xb32-600e_body7-640x640-dac2bf74_20231211.pth |
| - upstream SHA256: dac2bf749bbfb51e69ca577ca0327dff4433e3be9a56b782f0b7ef94fb45247e |
| - upstream paper: Lu et al., CVPR 2024 (arXiv:2312.07526) |
| - training set: body7 = COCO + AI Challenger + CrowdPose + MPII |
| + sub-JHMDB + Halpe + PoseTrack18 |
| |
| Conversion was performed by scripts/convert_rtmo.py in the acaua |
| repository. The conversion is deterministic and reversible: state |
| dict keys can be remapped back and are lossless (no quantization |
| or pruning applied). Training-only loss-module buffers |
| (head.loss_oks.sigmas, etc.) and data-preprocessor buffers |
| (data_preprocessor.mean, .std) are stripped — they have no role |
| at inference and are regenerated from config at training time. |
|
|
| Mirrored on 2026-04-22 by CondadosAI. |
|
|
| License |
| ------- |
| Licensed under the Apache License, Version 2.0 (the "License"); |
| you may not use this file except in compliance with the License. |
| You may obtain a copy of the License at |
|
|
| http://www.apache.org/licenses/LICENSE-2.0 |
|
|
| Unless required by applicable law or agreed to in writing, software |
| distributed under the License is distributed on an "AS IS" BASIS, |
| WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or |
| implied. See the License for the specific language governing |
| permissions and limitations under the License. |
|
|
