docs: add NOTICE attribution per Apache-2.0 s4(d) (code AND weights)

NOTICE

@@ -0,0 +1,66 @@
+UniFormer-S COCO top-down pose - acaua mirror (pure-PyTorch port)
+==================================================================
+
+This product includes:
+
+1. PORTED SOURCE CODE: a pure-PyTorch port of the UniFormer-pose
+   architecture (located in the acaua repository at
+   src/acaua/adapters/uniformer/pose/) is a derivative work of:
+
+     - Sense-X/UniFormer (Apache-2.0)
+       https://github.com/Sense-X/UniFormer @ main
+       Files derived from:
+         pose_estimation/mmpose/models/backbones/uniformer.py
+           (the global-attention UniFormer backbone, parameter-tree
+            identical to the detection variant — we reuse acaua's
+            UniFormer2DDense with hybrid=False, windows=False)
+         pose_estimation/mmpose/models/keypoint_heads/top_down_simple_head.py
+           (TopDownSimpleHead: 3x ConvTranspose2d-stride-2 + BN + ReLU
+            + 1x1 final conv)
+         pose_estimation/mmpose/core/evaluation/top_down_eval.py
+           (keypoints_from_heatmaps with post_process='default' branch:
+            argmax + sign(neighbor_diff) * 0.25 shift)
+
+2. CONVERTED WEIGHTS: the model.safetensors file in this mirror is a
+   key-remapped conversion of the upstream pretrained checkpoint:
+
+     - upstream repo:      Sense-X/UniFormer (Apache-2.0)
+     - upstream weights:   Google Drive file id
+                           162R0JuTpf3gpLe1IK6oxRoQK7JSj4ylx
+     - upstream filename:  top_down_256x192_global_small.pth
+     - upstream SHA256:    d77059e3e9322c0e20dc89dc0cf2a583ffe2ced7d3e9b350233738add570bc30
+     - upstream paper:     Li et al., "UniFormer: Unifying Convolution
+                           and Self-attention for Visual Recognition",
+                           arXiv:2201.09450 (ICLR 2022)
+
+   Key remap rule (eager, at conversion time):
+       backbone.*       ->  backbone.*       (no change)
+       keypoint_head.*  ->  head.*
+
+   Conversion was performed by scripts/convert_uniformer_pose.py in the
+   acaua repository. The adapter loads the safetensors file under
+   load_state_dict(strict=True) — key-naming drift between upstream and
+   the port is a hard error at load time.
+
+3. BUNDLED DETECTOR: this mirror's adapter loads
+   CondadosAI/rtmdet_t_coco at adapter-init time so callers get the
+   full top-down pipeline (detect persons -> per-person crop -> pose ->
+   inverse-warp) from a single predict(image) call. RTMDet-tiny carries
+   its own attribution chain — see CondadosAI/rtmdet_t_coco's NOTICE
+   file.
+
+Mirrored on 2026-04-25 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.