Parking Lot Yolo26m
This repository contains a YOLO detection model for parking-space occupancy detection. It predicts two classes:
Model source
- Kind
- Other
- File size
- 42.0 MB
- Version
- vRevision 0aa3b29
- Updated
- Sep 16, 2026
- Downloads
- 24
- Source
- Hugging Face
Source description
Parking Lot Occupancy Detection YOLO26m
This repository contains a YOLO detection model for parking-space occupancy detection. It predicts two classes:
space-emptyspace-occupied
The model is intended for aerial, elevated, or fixed-camera parking lot scenes where each parking space can be localized as a detection box.
Files
weights/parking_lot.pt: Ultralytics-compatible PyTorch checkpoint.metrics.json: validation metrics and example prediction settings.examples/predictions_imgsz1920/: example detections on parking lot images. These examples preserve each input image resolution and do not include any company logo or branding.
Validation Metrics
Internal validation split:
| Metric | Value |
|---|---|
| Images | 1,965 |
| Instances | 83,603 |
| Precision | 0.988 |
| Recall | 0.984 |
| [email protected] | 0.993 |
| [email protected]:0.95 | 0.954 |
Example Output
Green boxes indicate space-empty; orange boxes indicate space-occupied.
Prediction sample
Prediction contact sheet
Usage
from ultralytics import YOLO
model = YOLO("weights/parking_lot.pt")
results = model.predict("image.jpg", conf=0.25, imgsz=1920)
For dense small parking spaces, a larger imgsz such as 1280 or 1920 can improve visual coverage compared with low-resolution inference. Tiled inference can also be evaluated for high-resolution fixed cameras, but it may increase duplicate detections if post-processing is not tuned.
Feedback Wanted
Feedback and test results are welcome in the repository Discussions:
https://huggingface.co/simahanyan/parking-lot-yolo26m/discussions
Useful feedback includes missed detections in far/edge regions, duplicated boxes, angled or vertical parking rows, occlusion by trees, night/rain/snow scenes, and suggestions for training data, annotation policy, post-processing, or inference settings.
When sharing results, please include the input image/video frame if possible, the imgsz, confidence threshold, and whether tiled inference was enabled.
Limitations
- This is a rectangular bounding-box detector, not a calibrated parking-slot ROI system.
- For production parking management, fixed slot polygons/IDs and site calibration are recommended on top of detector output.
- Public example images are for visual smoke testing and should not be treated as an independent benchmark.
- Performance may vary under night scenes, rain/snow reflections, severe occlusion, unusual camera angles, and non-standard parking markings.
License and Use
License: other.
The checkpoint is shared for evaluation and feedback. Contact the model owner before commercial redistribution or deployment.
Checkpoint
- SHA256:
9b9c9185ce49cb71610341649ee49dd5ab4422bd9d5db0e818f61614835d4cae - Size:
44028761bytes
Sources
1 source
Sources
1 sourceVerified Sep 2
Model artifacts
1 artifact
Model artifacts
1 artifactSource excerpts
2 excerpts
Source excerpts
2 excerpts--- license: other library_name: ultralytics pipeline_tag: object-detection tags: - yolo - yolo26 - ultralytics - object-detection - parking - parking-lot - parking-space - occupancy-detection metrics: - precision - recall - map --- # Parking Lot Occupancy Detection YOLO26m This repository contains a YOLO detection model for parking-space occupancy detection. It predicts two classes: - `space-empty` - `space-occupied` The model is intended for aerial, elevated, or fixed-camera parking lot scenes where each parking space can be localized as a detection box. ## Files - `weights/parking_lot.pt`: Ultralytics-compatible PyTorch checkpoint. - `metrics.json`: validation metrics and example prediction settings. - `examples/predictions_imgsz1920/`: example detections on parking lot images. These examples preserve each input image resolution and do not include any company logo or branding. ## Validation Metrics Internal validation split: | Metric | Value | |---|---:| | Images | 1,965 | | Instances | 83,603 | | Precision | 0.988 | | Recall | 0.984 | | [email protected] | 0.993 | | [email protected]:0.95 | 0.954 | ## Example Output Green boxes indicate `space-empty`; orange boxes indicate `space-occupied`.   ## Usage ```python from ultralytics import YOLO model = YOLO("weights/parking_lot.pt") results = model.predict("image.jpg", conf=0.25, imgsz=1920) ``` For dense small parking spaces, a larger `imgsz` such as `1280` or `1920` can improve visual coverage compared with low-resolution inference. Tiled inference can also be evaluated for high-resolution fixed cameras, but it may increase duplicate detections if post-processing is not tuned. ## Feedback Wanted Feedback and test results are welcome in the repository Discussions: https://huggingface.co/simahanyan/parking-lot-yolo26m/discussions Useful feedback includes missed detections in far/edge regions, duplicated boxes, angled or vertical parking rows, occlusion by trees, night/rain/snow scenes, and suggestions for training data, annotation policy, post-processing, or inference settings. When sharing results, please include the input image/video frame if possible, the `imgsz`, confidence threshold, and whether tiled inference...
Source context: 24 downloads · 2 likes · Pipeline object-detection · Library ultralytics · Repo simahanyan/parking-lot-yolo26m
