Release reviewed experimental NCL candidate at 750 updates

Add the reviewed 750-update state dictionary and aggregate metadata, preserve the 375-update version, and document detection improvements, quantification limitations and upstream BSD-3-Clause attribution.

LICENSE.cellpose

@@ -0,0 +1,8 @@
+Copyright © 2020 Howard Hughes Medical Institute
+
+Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+
+Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+Neither the name of HHMI nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

NOTICE.md

@@ -0,0 +1,18 @@
+# Upstream model attribution
+
+The Cytellect NCL candidate is a fine-tune of the Cellpose-SAM `cpsam_v2`
+checkpoint distributed by MouseLand.
+
+- Upstream repository: https://huggingface.co/mouseland/cellpose-sam
+- Upstream revision: `7c61431b5fbb078f3296754bd15d9f51b320f837`
+- Upstream checkpoint SHA256: `0f1cc3f7ecdd8a037a57c6c48d9d8921391be4cbce3fa9f13c3e3a2e1253c667`
+- That revision's model card declares `bsd-3-clause`:
+  https://huggingface.co/mouseland/cellpose-sam/blob/7c61431b5fbb078f3296754bd15d9f51b320f837/README.md
+- The upstream copyright, conditions and disclaimer are reproduced in
+  `LICENSE.cellpose`.
+
+The fine-tuning changes the model weights. It does not imply endorsement by
+Howard Hughes Medical Institute or the upstream contributors.
+
+The release's model card and version metadata identify the exact fine-tuned
+checkpoint, its supported input conditions, and its evaluation limitations.

README.md

@@ -1,21 +1,45 @@
 ---
+license: bsd-3-clause
 tags:
 - microscopy
 - image-segmentation
 - experimental
 ---
-# NCL development pilot, 375 real-image updates
+# NCL development pilot
 
-Experimental NCL-staining-specific nucleolar segmentation weights, fine-tuned from an untouched cpsam_v2 checkpoint. Not adopted into the product. Not validated for NPM1 staining or cytoplasm. No synthetic images were used in this fine-tuning run.
+Experimental NCL-staining-specific nucleolar segmentation weights, fine-tuned from the untouched Cellpose-SAM `cpsam_v2` checkpoint. These weights are not the product default. Product integration and a user-operated local PoC are not completed. NPM1 staining and cytoplasm performance are not validated.
 
-Training used 8 approved objects in 2 regions from one experiment. Development used 4 reference objects in one region from a separately cultured and treated experiment; it was never a training input. There is no completed untouched final test.
+The 750-update candidate is intended for a bounded detection PoC. The earlier 375-update candidate remains available for reproducibility. This is not a general-purpose or fully validated nucleolar model. No synthetic images were used in this fine-tuning run.
 
-Under identical local product processing, development instance F1 changed from 0.5455 to 0.8000 (TP 3 to 4, FP 4 to 2, FN 1 to 0; matching IoU at least 0.5). Extra split fragments and individual boundary/area errors remain. Raw candidate F1 is 0.3636; product postprocessing materially affects the result. Total area agreement does not establish accurate per-object quantification. Training-region geometry regressions remain versus baseline.
+## Data and comparison
 
-This small development result does not establish general performance, native fluorescence measurement accuracy, or reduced human correction burden. Two-human reference agreement is not established. Preserve the base model and evaluate this candidate before any use.
+Training used 8 human-approved objects in 2 regions from one experiment. Development used 4 reference objects in one region from a separately cultured and treated experiment. That development image was not a training input, but it was used for candidate selection and is not an untouched final test. There is no completed untouched final test.
 
-Only model weights and aggregate metadata are included. No research images, annotations, optimizer state, or specimen identifiers are included.
+The table uses the same development reference and fixed local product preprocessing, parent-nucleus gating and postprocessing. Matching requires IoU at least 0.5. Training-region results are not pooled into development performance.
 
-Checkpoint SHA256: `bc5ea7f25ce9a9f96e9ff510cee20587e1328ac834a0f7f82aee0bc06156d906`.
+| Product output | TP | FP | FN | Instance F1 | Union Dice | Total area difference |
+|---|---:|---:|---:|---:|---:|---:|
+| Untouched baseline | 3 | 4 | 1 | 0.5455 | 0.8534 | -20.38% |
+| 375 updates | 4 | 2 | 0 | 0.8000 | 0.9349 | -2.69% |
+| 750 updates | 4 | 0 | 0 | 1.0000 | 0.8951 | -11.70% |
 
-See `ncl/r1-step0375.json` for the fixed recipe, package versions, provenance hashes, and aggregate metrics. No additional license grant is asserted by this model card; applicable upstream terms require separate review before redistribution.
+At 750 updates, the three better-segmented objects have area errors of +5, -6 and +5 pixels. The remaining object has 360 predicted pixels versus 608 reference pixels: **40.8% undersegmentation**, with IoU 0.5638. F1 of 1 means four matches pass the selected threshold; it does not mean perfect boundaries or quantification. At IoU 0.7, instance F1 is 0.75.
+
+Raw model output at 750 updates has TP 3, FP 1, FN 1 and F1 0.75. The fourth object's raw IoU is about 0.428; fixed product postprocessing raises it above 0.5. Raw and product output are not interchangeable. Removing the two unmatched fragments seen at 375 improves instance detection but worsens overall area agreement and union Dice. Some training-region geometry errors and a downstream parent-gate exclusion remain.
+
+## Use and limits
+
+Use the matching NCL input conditions and reviewed product processing. Preserve the base model and record the exact checkpoint identity. Do not treat this pilot's development scores as expected accuracy on new experiments. Human reference agreement between two annotators, native fluorescence measurement accuracy, reduced correction time and broad generalization are not established.
+
+Only model state-dict weights and aggregate metadata are distributed. No research images, annotations, optimizer state, raw logs, specimen identifiers or private filesystem paths are included.
+
+| Version | File | SHA256 |
+|---|---|---|
+| 750 updates | `ncl/r1-step0750.pt` | `9d2c2ec06aa1aa71f36695b49b0828f82fd3cb633da2832c99a9f194d0bd4099` |
+| 375 updates | `ncl/r1-step0375.pt` | `bc5ea7f25ce9a9f96e9ff510cee20587e1328ac834a0f7f82aee0bc06156d906` |
+
+The corresponding JSON files contain the fixed recipe, package versions, provenance hashes and aggregate metrics. These are state dictionaries; load only with a compatible, reviewed Cellpose-SAM implementation and `weights_only=True`, rather than executing arbitrary serialized objects.
+
+## License and attribution
+
+The upstream fixed-revision model card declares BSD-3-Clause. See `NOTICE.md` for the exact upstream revision and base checkpoint, and `LICENSE.cellpose` for the upstream copyright, conditions and disclaimer. The weights have been modified by fine-tuning; no upstream endorsement is implied.

ncl/r1-step0750.json

@@ -0,0 +1,68 @@
+{
+  "schema_version": 1,
+  "task": "NCL-stained nucleolar instance segmentation",
+  "status": "experimental_development_candidate_not_adopted",
+  "completed_updates": 750,
+  "checkpoint_sha256": "9d2c2ec06aa1aa71f36695b49b0828f82fd3cb633da2832c99a9f194d0bd4099",
+  "base_checkpoint_sha256": "0f1cc3f7ecdd8a037a57c6c48d9d8921391be4cbce3fa9f13c3e3a2e1253c667",
+  "data_manifest_sha256": "0ae2728c8e85ed94d55ca0baeb54bd72c139aac01a142d7ca6ebe4e8a5a5b7fe",
+  "training_code_sha256": "4d7797c716d9fde458b29ed3b1d07e8b990cbf826560d540a62cf5c0fb94d53a",
+  "inference_code_sha256": "c4b32c7910b3e00804b48794791468dfefedec7896ec770548f34872b4fd9453",
+  "recipe": {
+    "optimizer": "AdamW",
+    "lr_max": 1e-05,
+    "weight_decay": 0.1,
+    "warmup_updates": 100,
+    "batch_size": 1,
+    "rdrop": 0,
+    "augmentation": false
+  },
+  "environment": {
+    "cellpose": "4.2.1.1",
+    "torch": "2.8.0+cpu",
+    "numpy": "2.2.6",
+    "scipy": "1.16.2"
+  },
+  "evaluation_summary": {
+    "training_experiments": 1,
+    "training_regions": 2,
+    "training_objects": 8,
+    "development_experiments": 1,
+    "development_regions": 1,
+    "development_objects": 4,
+    "final_test_regions": 0,
+    "matched_product_baseline": {
+      "TP": 3,
+      "FP": 4,
+      "FN": 1,
+      "F1": 0.5454545454545454,
+      "union_Dice": 0.8534045393858478,
+      "total_area_relative_difference": -0.2038369304556355
+    },
+    "candidate_product_output": {
+      "TP": 4,
+      "FP": 0,
+      "FN": 0,
+      "F1": 1.0,
+      "union_Dice": 0.8950585838003057,
+      "total_area_relative_difference": -0.1170263788968825
+    },
+    "candidate_raw_output": {
+      "TP": 3,
+      "FP": 1,
+      "FN": 1,
+      "F1": 0.75,
+      "union_Dice": 0.8709005726184279,
+      "total_area_relative_difference": -0.15731414868105514
+    },
+    "limitations": [
+      "Development-only pilot; not a generalization or final validation claim.",
+      "Product output includes unchanged parent-nucleus and postprocessing stages; raw model output differs.",
+      "One development object is undersegmented by 40.8 percent in area despite passing the instance matching threshold.",
+      "Compared with 375 updates, fewer unmatched fragments but worse union Dice and total area agreement.",
+      "Training images are not independent validation; final training-output misses and geometry errors remain.",
+      "Two-human reference agreement, native fluorescence accuracy, and correction burden not measured.",
+      "Not validated for NPM1 staining or cytoplasm; product integration and actual user PoC not completed."
+    ]
+  }
+}

ncl/r1-step0750.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d2c2ec06aa1aa71f36695b49b0828f82fd3cb633da2832c99a9f194d0bd4099
+size 1233577155