Spaces:
Running on Zero
Running on Zero
Commit History
-
Add beginner and agronomist labeling guide Running
Verified signatureb1eacbf -
Explain browser GPU quota limits in the main result
Verified signatureb124366 -
Simplify leaf selection and explain model results
Verified signaturead3b6c5 -
Retry native downloads through the official channel
Verified signature421e7b7 -
Preserve UTF-8 labels in Space publication
Verified signature4836e42 -
Build SAM2 attention caches on CPU for ZeroGPU
Verified signaturebbb50de -
Normalize Space build files to Unix line endings
Verified signature74c90a7 -
Add reviewed paper methods and honest screening
Verified signatureIntegrate strict SAM2.1, SAM3, SCOLD, numeric-only SWAT, ImageJ and pliman adapters. Preserve advanced review and exports; correct conflicting verdicts and severity units. Comparisons remain unvalidated on the owner dataset until independent agronomist labels exist.bd7fc40
-
Preserve failed detector status in proposal mode
Verified signaturePropagate MODEL_ERROR separately from NO_LEAF_DETECTED when proposal generation fails. Both cases remain unclassified and require a selected region or one-leaf mode.5ef5d44 -
Protect whole-photo screening and candidate abstention
Verified signatureCheck that one-leaf mode uses the full image without running the leaf detector, and that multiple-leaf proposals never reach disease models. Protect the shared confirmation gate and ensure automatic batch exports have no binary predictions until a region is confirmed. Retain existing disagreement, mask and export checks.643a24d -
Stop diagnosing unreviewed detector proposals
Verified signatureDefault quick screening to a user-declared one-leaf close-up and process the entire photo, without detector cropping. Multiple-leaf mode only shows proposals and requires region confirmation in Advanced tools. Enforce the confirmation boundary in shared inference and keep automatic batch candidates unclassified. Preserve the five leaf-analysis models, advanced tools, independent masks and explicit failures.2cbd8d3 -
Shorten GPU reservations for single-crop inference
Verified signatureReserve 30 seconds per model call instead of 90. All weights are preloaded and each call handles one bounded crop. The live pipeline ran all six models successfully on the first candidate, then unnecessarily rejected remaining short calls because the 135-second quota reservation exceeded the account balance. Keep explicit timeout/error handling and free ZeroGPU hardware unchanged.ea1bea5 -
Check damage-detector disagreement and empty results
Verified signatureExtend the automatic-pipeline check to six models. Protect uncertainty when a healthy classifier conflicts with positive experimental damage detections. Ensure empty detector results do not create a healthy class or scores, and retain all prior review and export checks.12487c3 -
Add experimental DECAFIA damage localisation
Verified signatureRun the pinned CC BY 4.0 clean YOLOv8m checkpoint as a separate second opinion for rust, leaf miner and weevil defoliation. Draw cyan damage boxes, retain the simple upload/check flow and provide a separate advanced action. Preserve classifier outputs, flag positive-detector versus healthy-classifier disagreement as uncertain, and never turn no detections into healthy. Include source attribution and domain limits; no new dependencies.bc57bef -
Check the automatic leaf pipeline and review gates
Verified signatureExercise all five adapters for multiple leaves, preserve unreviewed automatic measurements and manual confirmation gates, and prevent healthy fallback after detector failures. No model downloads or GPU required.db05f6f -
Add a simple automatic coffee leaf check
Verified signatureUpload one photo and run the available five-model pipeline for every detected leaf. Show healthy-looking, sick-looking or uncertain results first. Preserve the advanced tools behind a closed accordion and keep automatic measurements explicitly unreviewed.5237d14 -
Add a regression check for leaf review and exports
Verified signature772d2fe -
Expose rust severity measurements in ledger CSV
Verified signatureAdd direct predicted and rust-reference severity columns for agreement analysis. Leave missing measurements blank and keep all-defect reference masks separate from rust reference area. Correct the batch device guidance. Validated with test_app.py under Gradio 6.28.0.5815bb1 -
Update requirements.txt
Verified signature2e4f334 -
Update requirements.txt
Verified signaturee7064a7 -
Update requirements.txt
Verified signature125cb90 -
Update app.py
Verified signature73db8a0 -
Update README.md
Verified signatureb4e5f1a -
Update README.md
Verified signature346c145 -
Update packages.txt
Verified signature625cdb2 -
Update requirements.txt
Verified signature40aaef1 -
Update app.py
Verified signaturea54839c -
Update app.py
Verified signature47f69ac -
Update app.py
Verified signature33f5b21 -
Update app.py
Verified signature5433414 -
Update app.py
Verified signaturec4a0f9d -
Update app.py
Verified signaturebb628cb -
Update requirements.txt
Verified signature3969de6 -
Update requirements.txt
Verified signature90f44d2 -
Update app.py
Verified signature4e5c03b -
Create packages.txt
Verified signatureea4c0ac -
Update requirements.txt
Verified signature0408972 -
Update app.py
Verified signatureeb23771 -
Update requirements.txt
Verified signature9bc2e3f -
Update requirements.txt
Verified signature7218398 -
Update app.py
Verified signature6ed0395 -
Create requirements.txt
Verified signature1ff0fb0 -
Update app.py
Verified signature240982d -
Update app.py
Verified signature4eed77d -
Create app.py
Verified signature4153a55 -
initial commit
Verified signature8ddd1ec