Spaces:
Sleeping
Sleeping
Commit History
-
Update main.py
Verified signature3a67dd1 -
Update main.py
Verified signature7994462
-
Update train.py
Verified signaturepoint correct .yaml path to the server8385446 -
Update train.py
Verified signatureupdated train.py due to error43e5db7 -
upload train.py
Verified signaturenew training script for cloud training with google colab ... dataset already uploaded to google drive0181bef -
Upload split.py
Verified signatureSplit dataset for yolo26m.pt77fe907
-
Upload 2 files
Verified signatureb455adb
-
Upload train.py
Verified signatureStart Training with 45000 images334fadc -
Upload verify.py
Verified signatureverifying labelled images1028208 -
Upload auto_label.py
Verified signaturelabel 45000 images using newely trained brain28d91cf -
Upload best.pt
Verified signatureNew Brain with Blast, Brown spot and Sheath Blight5ae39ff -
Upload train.py
Verified signaturetrain newely labelled images using new python scripte8fd3cd -
Upload split.py
Verified signaturesplit into train and val folders429f4df -
Upload augment_roboflow.py
Verified signatureuse augmentation for newely labelled imagesa677b76
-
Upload split.py
Verified signatureSplit newly labeled 400 images to train and val folders.ce79df3 -
Upload auto_label.py
Verified signatureAuto Label images using previously trained Ai Brainf89eb88 -
Testing new brain
Verified signatureTesting paddy disease images With new brain4d610f5 -
Train with new dataset
Verified signatureTrain with new datasetba8f90d
-
Update dataset splitting script for the 900-image dataset
Verified signatureThis script organizes the newly augmented 900-image dataset into the standard YOLO directory structure (train and val). Randomly shuffles and splits the data using an 80% Training / 20% Validation ratio. Safely maps and copies both images (.jpg, .png, .jpeg) and their corresponding .txt segmentation labels to their respective destination folders. Includes error handling to flag and log any missing label files during the transfer process to ensure dataset integrity.c51e7bb -
Update augmentation script to scale Teacher dataset to 900 images
Verified signatureScaled the base dataset target from 100 to 300 images per class (Blast, Brown Spot, Sheath Blight) to resolve underfitting and poor accuracy (35%) in the initial Teacher model. Retained geometric transformations (horizontal/vertical flips) with precise YOLO polygon coordinate inversions. Retained photometric transformations (Gaussian noise and brightness scaling) to improve model robustness against varied environmental drone lighting.96b923d -
Add YOLO medium segmentation training script for Teacher model
Verified signatureThis script initializes and trains the baseline "Teacher" model using the balanced 400-image dataset. Loads the YOLO Medium Segmentation architecture (yolo26m-seg.pt). Configures training for 200 epochs with an image resolution of 640px and a batch size of 4. Targets the teacher_v2.yaml dataset configuration to generate the foundational best.pt weights, which will later be used to auto-label the larger 45,000-image datasetd6c6958 -
Add YOLO dataset configuration for Teacher model training
Verified signatureThis YAML file configures the data paths and class mappings for training the 400-image "Teacher" segmentation model. Defines directory paths for the augmented training and validation image splits. Maps class IDs to their respective paddy diseases (0: blast, 1: brown_spot, 2: sheath_blight). Intentionally excludes the "Healthy" class, forcing YOLO to process empty .txt labels as background (negative) samples, which actively trains the model to reduce false positive detections.c37d176 -
Add script to split dataset into train and validation sets
Verified signatureThis script organizes the balanced seed dataset into the required YOLO folder structure (train and val). Randomly shuffles and splits the dataset using an 80% Training / 20% Validation ratio. Automatically routes both images and their corresponding .txt segmentation labels. Includes a safety fallback to generate empty label files for background/healthy images, preventing YOLO training errors.5bbbbc0 -
Add script to balance and augment paddy dataset to 400 images
Verified signatureThis script resolves class imbalance by dynamically augmenting the dataset until it reaches exactly 100 images per class (Blast, Brown Spot, Sheath Blight, Healthy). Uses OpenCV to apply random transformations (h_flip, v_flip, brightness, noise). Automatically calculates and shifts YOLO segmentation polygon coordinates to match the newly generated images.b730b20
-
Update main.py
Verified signature147d1d7 -
Update main.py
Verified signaturef6f1d5f -
Update main.py
Verified signature01efda8 -
Update main.py
Verified signatureb00dada
-
Update main.py
Verified signature27c1847 -
Update main.py
Verified signatureccd2742 -
Update requirements.txt
Verified signatureee6cdb6 -
Update Dockerfile
Verified signature85f8047 -
Update requirements.txt
Verified signature4794afe -
Update Dockerfile
Verified signature197fe55 -
Delete packages.txt
Verified signaturea47e8dc -
Update requirements.txt
Verified signature902bc51 -
Create packages.txt
Verified signatureaf0fab7 -
Delete packages.txt
Verified signaturef6be40f -
Create packages.txt
Verified signature0bf9d9a -
Update requirements.txt
Verified signature9986fac -
Update main.py
Verified signaturec4a6183 -
Update Dockerfile
Verified signature7ea6abe -
Upload 4 files
Verified signature160521d -
initial commit
Verified signature0fc5652