Single-site label, train, and scan â all in one place
đ Dataset
View dataset and filter by SKU
đˇī¸ SKU Manager
Manage cropped images by SKU
đ Label
Annotate images with bounding boxes + 2-tier hierarchy
đ Scan
Analyze dataset and label quality
đ Train
Full training or fine-tuning
đž Backups
Manage model backups and history
⥠FAISS Index
Generate, upload, and manage FAISS index
đ Dataset Manager - Raw Images
Manage all user-uploaded images with file explorer, preview, and delete options
đ Total Images
0
đ Location
dataset/images/
đ Files(0)
Loading files...
âŗ Uploading...
đˇī¸ SKU Manager - Cropped Images
Manage cropped images organized by SKU folder
đĻ Bulk Import SKUs from ZIP
Upload a data.zip file with structure: data/{SKU_NAME}/{images}
No file selected
âŗ Uploading and extracting...
đĻ Total SKUs
0
đ Location
openclip_dataset/
SKU Name
Images
Size (MB)
Actions
Loading SKU list...
Loading images...
đ Total: 0
đ Files(0)
Loading files...
Uploading...
đ Label Products (SKU Assignment)
đ¯ Object Detection
đĻ product
Single class detection model. Draw boxes around products.
#
SKU ID
Action
No SKUs added yet
#
Position
Assign SKU
Action
No boxes detected
Select label for new box
đ 2-Tier Inference & Detection
đī¸ Upload Image for Detection
Upload an image to detect products and match them to known SKUs
Processing...
đ Detection Results
Products Detected
0
Matched SKUs
0
âšī¸
Matching Mode:
No SKU matching
Uploaded Image:
Detected Products:
ID
Confidence
Bounding Box (x,y,w,h)
Matched SKU
SKU Similarity
SKU Matches Summary:
đ¤ OCR Keywords (High Confidence > 50%):
#
Keyword
Confidence
⨠SKU Match from OCR Keywords
đŧī¸ Detected Crops:
đ Unified Detection Results
đ¯ Detection Summary
0 Products0 SKUs
Total Detections
0
Image Dimensions
-
Matched SKUs
0
đĻ Detected Products Breakdown
-
đ Raw JSON (Click to expand)
{}
đ Model Training
đ¤ Upload Trained Model as scan_model.pt
Upload your trained .pt file to use it for scanning. It will be saved as scan_models/scan_model.pt
No scan model uploaded yet
âŗ Uploading scan model...
Training Parameters
Training cycles
Input image size
Batch size
Compute device
Early stopping patience
đ Training Info:
âĸ Train from Start: Uses base HOLO11 model, better for new datasets
âĸ Fine-tune: Uses existing trained model, faster convergence
âĸ Dataset: 18 images with 40 annotations
âĸ Augmentation: Mosaic, mixup, color jitter enabled
âĸ Auto-backup: Models saved at each epoch
âĸ Early stopping: Prevents overfitting Current Config:100ep à 640px à 8batch
⥠FAISS Index Management
Generate, manage, upload, and download FAISS index for fast SKU matching (15-70x speedup)
đ Status
Not loaded
đĻ SKUs
--
đž Size
-- MB
âī¸ Index Operations
âŗ Processing...
đž Model Backups
Filename
Size (MB)
Modified
Action
Loading backups...
đ¤ Upload FAISS Index
Upload a previously saved FAISS index file. You need BOTH files: