The ability of computer vision applications to address various business challenges makes it one of the most exciting technologies available today. Computer vision use cases in agriculture are in high demand thanks to the many applicable solutions the new technology provides.
From weed detection and crop disease treatment to automated spraying via drones and autonomous tractors, the application of machine vision and image processing techniques in agriculture are endless.
This post will help you discover six of the best open source datasets for computer vision and image processing in the agriculture industry to optimize productivity, boost yield, decrease costs, and increase profits.
In addition to these curated agriculture datasets, thousands more datasets are available for free on Roboflow Universe.
1. Weeds Computer Vision Project
Link: https://universe.roboflow.com/augmented-startups/weeds-nxe1w
Project Type: Object-Detection
Subject: Plants
Classes: weeds
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The weeds computer vision project is a collection of 3.7K training, 359 validation and 180 garden weeds images that can be used with YOLOR for object detection to detect weeds in complex backgrounds and surroundings. This dataset can train a machine vision algorithm capable of detecting specific plants. You can reduce labour costs and maximize yields by automating the weeding process or using precision agriculture to reduce fertilizer use.
Test the model's performance by calling Roboflow's API on the images.
2. Apple Sorting Computer Vision Project
Link: https://universe.roboflow.com/arfiani-nur-sayidah-9lizr/apple-sorting-2bfhk
Project Type: Object-Detection
Subject: Apple
Classes: apple, damaged_apple
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
Apple sorting computer vision project is a dataset of 253 training, 103 validation, and 5 testing images of apples. The dataset can be used for sorting store quality apples from damaged apples. The dataset has a balanced set of classes, i.e. apples (2152 classes) and damaged_apples (708 classes). The dataset can train an object detection model capable of sorting only high quality apple products and sorting out defective apple products for quality inspection systems. It can also count the total number of apple products and grade them into various grades like premium, second, peeler, cull, etc
Test the model's performance by calling Roboflow's API pretrained on the images.
3. Chicken Detection and Tracking Computer Vision Project
Link: https://universe.roboflow.com/chickens/chicken-detection-and-tracking
Project Type: Object-Detection & Tracking
Subject: Chicken
Classes: chicken
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The chicken detection and tracking computer vision project is a dataset of 128 training, 38 validation and 19 testing images of chicken. The dataset can be used to train an object detection model that can identify chickens and perform object-tracking on chickens. The model can also be used for real-time detection and tracking of the activity of laying hens in poultry farms and counting chickens to give them self-identification numbers.
Test the model's performance by calling Roboflow's API pretrained on the images.
4. Honey Bee Detection Model Computer Vision Project
Link: https://universe.roboflow.com/matt-nudi/honey-bee-detection-model-zgjnb
Project Type: Object-Detection
Subject: Workers, Drones, Queens, Pollenbees
Classes: bee, drone, pollenbee, queen
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The honey bee detection model is a dataset of 4.3K training, 176 validation and 122 testing images of classes including bee, drones, pollen bee, and queen. The dataset was collected for real-time counting of bees entering the hive but later was extended to count and detect drones, queens, and pollen bees. The dataset can also be used for correlating behaviors at the hive's entrance with the weather, temperature, surroundings, etc.
Test the model's performance by calling Roboflow's API pretrained on the images.
5. Aerial Sheep Computer Vision Project
Link: https://universe.roboflow.com/riis/aerial-sheep
Project Type: Object-Detection
Subject: Sheep
Classes: sheep
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The aerial sheep computer vision project is a dataset containing 3.6K training, 350 validation, and 174 testing images of sheep. The dataset contains images taken from a drone with instances of sheep in them. One thing to mention is that the sheep do not differentiate between gender or breed. It can also be used for counting sheep, keeping track of sheep in the yard, and detecting sheep in an image.
Test the model's performance by calling Roboflow's API pretrained on the images.
6. Detecting Diseases Computer Vision Project
Link: https://universe.roboflow.com/artificial-intelligence-82oex/detecting-diseases
Project Type: Object-Detection
Subject: Disease
Classes: 10-515625, ALS, Angular Leafspot, Anthracnose Fruit Rot, Bean Rust, Blossom Blight, Gray Mold, Leaf Spot, Powdery Mildew Fruit, Powdery Mildew Leaf, disease, leaf mold, spider mites
Download Formats: YOLOv5, YOLOv7, MT-YOLOv6, COCO JSON, YOLO Darknet, Pascal VOC XML, TFRecord, CreateML JSON, etc
The detecting diseases computer vision project is a dataset of 2.9K training, 1.4K validation and 1.2K testing images of all the different classes of diseases usually seen in plants. The project can be utilized to train a model that can categorize if there are diseases and pests in plants or crops. This model can help reduce manual monitoring in crops and improve the yield of plants by detecting diseases. A total of 5494 images have pre-processing applied to each image.
Using Open Source Agriculture Datasets for Computer Vision
In the last decade, there has been skyrocketing interest in sustainability in the agricultural industry and machine vision is a critical technology to improve efficiency in agriculture.
Data collected through embedded devices throughout farms can be used for quality assurance and product inspection, monitoring for safety and compliance, augmenting manual processes, precision agriculture, inventory management, and anomaly detection.
Sign up for a free Roboflow account to begin adding images from open source datasets and then train a computer vision model in one-click. You'll have a fully running computer vision pipeline in no time.