vtrace-behavior 1.0.0__tar.gz

vtrace_behavior-1.0.0/LICENSE

@@ -0,0 +1,176 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to the Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by the Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding any notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS

vtrace_behavior-1.0.0/MANIFEST.in

@@ -0,0 +1,24 @@
+# Config files (Python-based configs, included as package data)
+recursive-include configs *.py
+
+# Entry-point scripts
+recursive-include tools *.py
+
+# The V-TRACE GUI (a single static page, shipped as-is)
+recursive-include vtrace/static *
+
+# CalMS21 demo annotations (~50 KB). The videos are NOT here: they are
+# fetched from the dataset's own record at CaltechDATA.
+recursive-include vtrace/demo_data *.csv *.txt *.json
+
+# Documentation
+include README.md
+include LICENSE
+
+# Exclude dev/test files
+exclude .gitignore
+prune exps
+prune data
+prune pretrained
+prune figures
+prune literature_review

vtrace_behavior-1.0.0/PKG-INFO

@@ -0,0 +1,218 @@
+Metadata-Version: 2.4
+Name: vtrace-behavior
+Version: 1.0.0
+Summary: V-TRACE: continuous ethograms of animal behavior from untrimmed video
+License: Apache-2.0
+Project-URL: Homepage, https://github.com/KunmingS/TRACE
+Project-URL: Documentation, https://kunmings.github.io/TRACE/
+Project-URL: Repository, https://github.com/KunmingS/TRACE
+Project-URL: Issues, https://github.com/KunmingS/TRACE/issues
+Keywords: ethogram,animal-behavior,behavior-annotation,video-analysis,deep-learning,pytorch
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Scientific/Engineering :: Image Recognition
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: torch>=2.0.1
+Requires-Dist: torchvision>=0.15.2
+Requires-Dist: decord>=0.6.0
+Requires-Dist: einops>=0.6.0
+Requires-Dist: scipy>=1.9.0
+Requires-Dist: pandas>=1.5.0
+Requires-Dist: tqdm>=4.64.0
+Requires-Dist: numpy>=1.23.5
+Requires-Dist: opencv-python-headless>=4.6.0
+Requires-Dist: static-ffmpeg>=2.5
+Requires-Dist: rich>=13.0
+Requires-Dist: prompt_toolkit>=3.0
+Provides-Extra: dev
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: flake8; extra == "dev"
+Dynamic: license-file
+
+# V-TRACE — Video-based Temporal Recognition and Annotation of Continuous Ethograms of Animal Behavior
+
+V-TRACE turns untrimmed animal-behavior video into continuous ethograms: label a few
+recordings in the browser, train a detector on them, and have it annotate the rest.
+
+## Documentation
+
+[kunmings.github.io/TRACE](https://kunmings.github.io/TRACE/)
+
+## Install
+
+Create a Python 3.9+ environment first. Choose one:
+
+```bash
+conda create -n vtrace python=3.11 pip
+conda activate vtrace
+```
+
+```bash
+mamba create -n vtrace python=3.11 pip
+mamba activate vtrace
+```
+
+```bash
+uv venv --python 3.11 --seed .venv
+source .venv/bin/activate
+```
+
+Then install V-TRACE and download the model weights:
+
+```bash
+python -m pip install vtrace-behavior
+vtrace prepare --weights all
+```
+
+Training, evaluation and prediction need a CUDA-capable PyTorch environment.
+Annotation does not — the browser app works anywhere.
+
+## Usage
+
+Run it with no arguments:
+
+```bash
+vtrace
+```
+
+That opens an interactive session. The annotator starts with it, so the screen
+gives you its address — open `http://localhost:8765` in Chrome or Edge — and you
+type the rest of the commands at the `vtrace >` prompt, without repeating the
+program's name (`demo predict`, not `vtrace demo predict`). Tab completes,
+ctrl-r searches history, `exit` leaves.
+
+The annotator is a single static page. It reads videos and writes annotations
+**on the computer running the browser**, through the File System Access API:
+nothing is uploaded and there is no server-side state. Use it to open
+recordings, define behaviors, mark events on the timeline, and review model
+predictions; training, evaluation and prediction run from the command line.
+
+`vtrace app` serves the annotator on its own, without the session — for scripts,
+or when that is all you want.
+
+Every command below also works directly from a shell, prefixed with `vtrace`.
+
+## Try the demo
+
+A walkthrough on CalMS21 mouse social behavior, using the *complete* Task-1
+benchmark in its official split — 70 training videos (4.7 h, 2287 labelled
+bouts) and 19 test videos (2.4 h, 1521 bouts):
+
+```bash
+vtrace demo predict     # predict on a held-out video, write its predictions
+vtrace demo download    # fetch the videos (train ~19 GB, test ~11 GB)
+vtrace demo train       # prep + train on the 70 official training videos
+```
+
+V-TRACE does not re-host CalMS21. The annotation CSVs ship with the package; the
+videos come straight from the
+[official CalMS21 release](https://data.caltech.edu/records/s0vdx-0k302)
+(doi:10.22002/D1.1991), one member at a time over HTTP range requests, so
+nothing downloads the full 28 GB archive and an interrupted fetch resumes by
+skipping what is already on disk. `--split train` / `--split test` limits what
+is pulled, and `--from /path/to/task1_videos_mp4.zip` reads a copy you already
+have.
+
+`vtrace demo predict` pulls the single video it needs and runs in about a minute.
+`vtrace demo train` is a real training run on the full benchmark, not a smoke
+test: 337 iterations per epoch over ten epochs, roughly three hours on one
+modern GPU (~23 GB of VRAM).
+
+If you use the demo data, cite CalMS21: Sun et al., *The Multi-Agent Behavior
+Dataset: Mouse Dyadic Social Interactions*, NeurIPS 2021 Datasets & Benchmarks.
+
+`vtrace demo download` assembles an ordinary model directory in `~/.vtrace/demo`, so
+afterwards the plain commands work against it too:
+
+```bash
+vtrace predict --model-dir ~/.vtrace/demo/model --input /my/video.mp4
+```
+
+```bash
+# Check whether PyPI has a newer V-TRACE release
+vtrace update
+
+# Train from selected video/annotation pairs
+vtrace train --model maev2 --work-dir /my/dataset --pairs video01.mp4=video01_final.csv video02.mp4=video02.csv
+
+# Evaluate the training artifact, or evaluate on held-out video/annotation pairs
+vtrace eval --model-dir /my/dataset/model_YYYYMMDD_HHMMSS
+vtrace eval --model-dir /my/dataset/model_YYYYMMDD_HHMMSS --work-dir /my/testset --pairs video03.mp4=video03.csv
+
+# Predict on new videos and write annotation drafts
+vtrace predict --model-dir /my/dataset/model_YYYYMMDD_HHMMSS --input /path/to/video.mp4 --threshold 0.25
+
+# Run prep -> train -> predict end to end
+vtrace pipeline --train --infer --model maev2 \
+  --work-dir /my/dataset --pairs video01.mp4=video01_final.csv \
+  --input /path/to/new/videos
+```
+
+`--pairs` is explicit: each item is `VIDEO_PATH=CSV_PATH`. A source video can
+have multiple annotation CSVs beside it, such as `video01_draft.csv`,
+`video01_final.csv`, or a reviewed prediction exported from the annotator; each
+training or evaluation pair chooses the annotation file to use for that video.
+Relative paths are resolved against `--work-dir`, so
+`video01.mp4=video01_final.csv` means both files are inside the work directory.
+Training creates a self-contained `model_YYYYMMDD_HHMMSS/` folder under
+`--work-dir`.
+
+V-TRACE annotation CSVs are time-based:
+
+```csv
+labelId,timestamp,endTimestamp
+grooming,12.430,18.970
+rearing,42.100,45.650
+```
+
+Prediction writes **one file per video**, named after it and placed beside it:
+
+```
+my_video.mp4
+my_video.predict.json
+```
+
+Re-running prediction replaces that file rather than accumulating copies — a
+video has exactly one current prediction. The JSON holds the merged behaviour
+bouts with their confidences, in the shape the V-TRACE annotator imports, so the
+same folder opened in `vtrace app` shows the video with its predictions ready for
+review. `--output DIR` writes the files somewhere else, keeping the names.
+
+## Shipped research configs and checkpoints
+
+The repo carries the CalMS21 configurations behind the paper's results, and the
+matching checkpoints are hosted on the GitHub `weights` release. **No checkpoint
+is bundled in the repo or the wheel** — any config or `--checkpoint` argument
+that names a registered file downloads it on first use (SHA256-verified, cached
+under `~/.vtrace/pretrained`, override with `TRACE_WEIGHTS_DIR`):
+
+| Config | Checkpoint | What it is |
+|---|---|---|
+| `configs/calms21_distill_vmaeB.py` | `calms21_vitB_distilled_best.pth` | The CalMS21 detector: a VideoMAE V2 ViT-B student, frozen ViT plus trained per-block adapters and a dense per-frame head. This is what `vtrace demo predict` runs. |
+| `configs/calms21_vjepa2.py` | `calms21_vjepa2_teacher95.pth` | The V-JEPA2 ViT-L **encoder** fine-tuned on CalMS21 — the teacher the student above was distilled from. Encoder weights only, so it initialises a backbone rather than predicting on its own. |
+
+Two pretrained bases complete the distillation chain: `vitB_videomaev2_k400.pth`
+(the student's frozen ViT-B) and `vjepa2_vitl_encoder.pth` (the teacher's
+starting encoder, before CalMS21 fine-tuning). To reproduce the distillation
+itself, see `tools/distill_jepa_to_vmae.py` and `tools/bake_distill_adapters.py`.
+
+Example — score the CalMS21 detector on the 19 official test videos:
+
+```bash
+torchrun --nproc_per_node=1 tools/test.py configs/calms21_distill_vmaeB.py \
+    --checkpoint calms21_vitB_distilled_best.pth
+```
+
+## License
+
+Apache 2.0. See [LICENSE](LICENSE).

vtrace_behavior-1.0.0/README.md

@@ -0,0 +1,177 @@
+# V-TRACE — Video-based Temporal Recognition and Annotation of Continuous Ethograms of Animal Behavior
+
+V-TRACE turns untrimmed animal-behavior video into continuous ethograms: label a few
+recordings in the browser, train a detector on them, and have it annotate the rest.
+
+## Documentation
+
+[kunmings.github.io/TRACE](https://kunmings.github.io/TRACE/)
+
+## Install
+
+Create a Python 3.9+ environment first. Choose one:
+
+```bash
+conda create -n vtrace python=3.11 pip
+conda activate vtrace
+```
+
+```bash
+mamba create -n vtrace python=3.11 pip
+mamba activate vtrace
+```
+
+```bash
+uv venv --python 3.11 --seed .venv
+source .venv/bin/activate
+```
+
+Then install V-TRACE and download the model weights:
+
+```bash
+python -m pip install vtrace-behavior
+vtrace prepare --weights all
+```
+
+Training, evaluation and prediction need a CUDA-capable PyTorch environment.
+Annotation does not — the browser app works anywhere.
+
+## Usage
+
+Run it with no arguments:
+
+```bash
+vtrace
+```
+
+That opens an interactive session. The annotator starts with it, so the screen
+gives you its address — open `http://localhost:8765` in Chrome or Edge — and you
+type the rest of the commands at the `vtrace >` prompt, without repeating the
+program's name (`demo predict`, not `vtrace demo predict`). Tab completes,
+ctrl-r searches history, `exit` leaves.
+
+The annotator is a single static page. It reads videos and writes annotations
+**on the computer running the browser**, through the File System Access API:
+nothing is uploaded and there is no server-side state. Use it to open
+recordings, define behaviors, mark events on the timeline, and review model
+predictions; training, evaluation and prediction run from the command line.
+
+`vtrace app` serves the annotator on its own, without the session — for scripts,
+or when that is all you want.
+
+Every command below also works directly from a shell, prefixed with `vtrace`.
+
+## Try the demo
+
+A walkthrough on CalMS21 mouse social behavior, using the *complete* Task-1
+benchmark in its official split — 70 training videos (4.7 h, 2287 labelled
+bouts) and 19 test videos (2.4 h, 1521 bouts):
+
+```bash
+vtrace demo predict     # predict on a held-out video, write its predictions
+vtrace demo download    # fetch the videos (train ~19 GB, test ~11 GB)
+vtrace demo train       # prep + train on the 70 official training videos
+```
+
+V-TRACE does not re-host CalMS21. The annotation CSVs ship with the package; the
+videos come straight from the
+[official CalMS21 release](https://data.caltech.edu/records/s0vdx-0k302)
+(doi:10.22002/D1.1991), one member at a time over HTTP range requests, so
+nothing downloads the full 28 GB archive and an interrupted fetch resumes by
+skipping what is already on disk. `--split train` / `--split test` limits what
+is pulled, and `--from /path/to/task1_videos_mp4.zip` reads a copy you already
+have.
+
+`vtrace demo predict` pulls the single video it needs and runs in about a minute.
+`vtrace demo train` is a real training run on the full benchmark, not a smoke
+test: 337 iterations per epoch over ten epochs, roughly three hours on one
+modern GPU (~23 GB of VRAM).
+
+If you use the demo data, cite CalMS21: Sun et al., *The Multi-Agent Behavior
+Dataset: Mouse Dyadic Social Interactions*, NeurIPS 2021 Datasets & Benchmarks.
+
+`vtrace demo download` assembles an ordinary model directory in `~/.vtrace/demo`, so
+afterwards the plain commands work against it too:
+
+```bash
+vtrace predict --model-dir ~/.vtrace/demo/model --input /my/video.mp4
+```
+
+```bash
+# Check whether PyPI has a newer V-TRACE release
+vtrace update
+
+# Train from selected video/annotation pairs
+vtrace train --model maev2 --work-dir /my/dataset --pairs video01.mp4=video01_final.csv video02.mp4=video02.csv
+
+# Evaluate the training artifact, or evaluate on held-out video/annotation pairs
+vtrace eval --model-dir /my/dataset/model_YYYYMMDD_HHMMSS
+vtrace eval --model-dir /my/dataset/model_YYYYMMDD_HHMMSS --work-dir /my/testset --pairs video03.mp4=video03.csv
+
+# Predict on new videos and write annotation drafts
+vtrace predict --model-dir /my/dataset/model_YYYYMMDD_HHMMSS --input /path/to/video.mp4 --threshold 0.25
+
+# Run prep -> train -> predict end to end
+vtrace pipeline --train --infer --model maev2 \
+  --work-dir /my/dataset --pairs video01.mp4=video01_final.csv \
+  --input /path/to/new/videos
+```
+
+`--pairs` is explicit: each item is `VIDEO_PATH=CSV_PATH`. A source video can
+have multiple annotation CSVs beside it, such as `video01_draft.csv`,
+`video01_final.csv`, or a reviewed prediction exported from the annotator; each
+training or evaluation pair chooses the annotation file to use for that video.
+Relative paths are resolved against `--work-dir`, so
+`video01.mp4=video01_final.csv` means both files are inside the work directory.
+Training creates a self-contained `model_YYYYMMDD_HHMMSS/` folder under
+`--work-dir`.
+
+V-TRACE annotation CSVs are time-based:
+
+```csv
+labelId,timestamp,endTimestamp
+grooming,12.430,18.970
+rearing,42.100,45.650
+```
+
+Prediction writes **one file per video**, named after it and placed beside it:
+
+```
+my_video.mp4
+my_video.predict.json
+```
+
+Re-running prediction replaces that file rather than accumulating copies — a
+video has exactly one current prediction. The JSON holds the merged behaviour
+bouts with their confidences, in the shape the V-TRACE annotator imports, so the
+same folder opened in `vtrace app` shows the video with its predictions ready for
+review. `--output DIR` writes the files somewhere else, keeping the names.
+
+## Shipped research configs and checkpoints
+
+The repo carries the CalMS21 configurations behind the paper's results, and the
+matching checkpoints are hosted on the GitHub `weights` release. **No checkpoint
+is bundled in the repo or the wheel** — any config or `--checkpoint` argument
+that names a registered file downloads it on first use (SHA256-verified, cached
+under `~/.vtrace/pretrained`, override with `TRACE_WEIGHTS_DIR`):
+
+| Config | Checkpoint | What it is |
+|---|---|---|
+| `configs/calms21_distill_vmaeB.py` | `calms21_vitB_distilled_best.pth` | The CalMS21 detector: a VideoMAE V2 ViT-B student, frozen ViT plus trained per-block adapters and a dense per-frame head. This is what `vtrace demo predict` runs. |
+| `configs/calms21_vjepa2.py` | `calms21_vjepa2_teacher95.pth` | The V-JEPA2 ViT-L **encoder** fine-tuned on CalMS21 — the teacher the student above was distilled from. Encoder weights only, so it initialises a backbone rather than predicting on its own. |
+
+Two pretrained bases complete the distillation chain: `vitB_videomaev2_k400.pth`
+(the student's frozen ViT-B) and `vjepa2_vitl_encoder.pth` (the teacher's
+starting encoder, before CalMS21 fine-tuning). To reproduce the distillation
+itself, see `tools/distill_jepa_to_vmae.py` and `tools/bake_distill_adapters.py`.
+
+Example — score the CalMS21 detector on the 19 official test videos:
+
+```bash
+torchrun --nproc_per_node=1 tools/test.py configs/calms21_distill_vmaeB.py \
+    --checkpoint calms21_vitB_distilled_best.pth
+```
+
+## License
+
+Apache 2.0. See [LICENSE](LICENSE).