Fruit Recognition
Set of images of fruits and vegetables intended for visual recognition. The dataset contains 33 classes, with a standard size of 100x100 pixels.
22,495 JPG images (100x100px), spread over 33 classes
TJ NMLO Public Dataset (free use without restrictive mention)
Description
The dataset Fruit Recognition contains over 22,000 images of fruits and vegetables, each classified in one of 33 categories. The images are uniform in size (100x100 pixels) and include rotations to increase the robustness of the model. It is structured in two sets: a training set and a test set, each organized according to machine learning best practices.
What is this dataset for?
- Train an image classification model for fruits and vegetables
- Test simple or complex visual recognition algorithms
- Create educational or food applications based on food detection
Can it be enriched or improved?
Yes. It is possible to artificially increase the data via rotation, zoom or blur, or to add metadata (season, origin, dominant color). Manual or automatic segmentation can also improve precision on more complex tasks such as multi-object detection.
🔎 In summary
🧠 Recommended for
- Deep learning beginners
- Food AI projects
- Interactive classification demos
🔧 Compatible tools
- TensorFlow
- Keras
- PyTorch
- FastAI
- Teachable Machine
💡 Tip
To increase the robustness of your model, apply pretreatment of variable lighting and artificial backgrounds to simulate real conditions.
Frequently Asked Questions
Is the dataset suitable for a real-time recognition project?
Yes, its reduced image size (100x100) makes it very effective for lightweight models, suitable for mobile or embedded deployments.
Is the fruit well-balanced between classes?
Overall yes, although some classes may contain slightly more images. Prior analysis is recommended.
Can this dataset be used for commercial use?
Yes, the source is public and no explicit restrictions are indicated. However, it is prudent to check the terms every time you update.
