Object Detector
Detect 80 common object classes with YOLOv4-tiny.
What does the object detector do?
The freeimgtool object detector identifies the location and class of common objects in an uploaded photo. Behind the scenes it runs YOLOv4-tiny, a fast small variant of the popular YOLO family, loaded through OpenCV's DNN module. YOLOv4-tiny is trained on the COCO dataset, which means it can recognise 80 common everyday classes — people, cars, dogs, cats, food items, household objects, sports equipment, and more.
The result is both a JSON list of detections with confidence scores and an annotated image with coloured boxes drawn around each object. Adjust the confidence threshold to trade recall for precision. Lower the threshold to catch faint detections, raise it to keep only confident ones.
How to use it
- Click the upload box and pick a photo (up to 5 MB).
- Drag the confidence slider. Start around 40% for general photos.
- Click Detect objects.
- Read the table of detections.
- Download the annotated image with class labels and boxes.
What classes does it know?
YOLOv4-tiny with COCO weights knows 80 classes including: person, bicycle, car, motorbike, aeroplane, bus, train, truck, boat, traffic light, fire hydrant, bench, bird, cat, dog, horse, sheep, cow, elephant, bear, zebra, giraffe, backpack, umbrella, handbag, suitcase, sports ball, kite, baseball bat, skateboard, surfboard, tennis racket, bottle, wine glass, cup, fork, knife, spoon, bowl, banana, apple, sandwich, orange, broccoli, carrot, hot dog, pizza, donut, cake, chair, sofa, bed, dining table, toilet, TV, laptop, mouse, remote, keyboard, cell phone, microwave, oven, toaster, sink, refrigerator, book, clock, vase, scissors, teddy bear, hairdryer, toothbrush.
Why use this object detector
Useful for quick inventory checks: counting people in an event photo, spotting items in a desk photo, checking whether a packing list is on camera, building a teaching example for computer vision class. The model is fast and free to use.
Who uses it
Students learning computer vision. Real-estate agents cataloguing furniture in a room. Retailers spot-checking stock photos. Event organisers counting attendees. Researchers building datasets. Anyone curious about what a modern object detector sees in their photo.
Frequently asked questions
Is the object detector free?
Yes. No signup, no premium tier. Display ads support the site.
What is the maximum upload size?
Five megabytes per photo. The detector downscales internally to 1280 pixels on the long edge for speed.
Why is the first request slow?
YOLOv4-tiny weights are downloaded and cached on the server the first time the endpoint is called. Subsequent requests are fast.
Can it identify a specific person or brand?
No. It only labels generic classes from COCO (e.g. "person", "car", "bottle"). Identity, brand, or fine-grained classification is out of scope.
Why did it miss something obvious?
YOLOv4-tiny is the small variant — it trades accuracy for speed. Try lowering the confidence threshold. Heavier YOLO variants and Detectron-style models are on the roadmap for the upgraded server.
What output format do I get?
An annotated JPG with coloured bounding boxes and labels. The JSON response also includes class names, confidences, and box coordinates.
Is my photo stored?
Uploads are kept only long enough to return the annotated image link and are removed automatically. Privacy policy has the details.
Does it work on a phone?
Yes. The upload box opens the camera roll on most phones directly.