Enabling Efficient Wildlife Preservation with an AI based Solution for Camera Trap Image Analysis
Reduction in Image Analysis Time
Reduction in Manual Effort
INTRODUCTION
In the realm of wildlife preservation, the Forest Department of Karnataka stands out as a leader in using technology for conservation. One of its partners, Kimzuka Solutions, creates equipment like camera traps that monitor large forest areas, capturing many images of the wilderness. These camera traps, placed to capture any passing wildlife, store their images on SD cards. However, sorting through these images manually was laborious and time-consuming. This process was crucial for counting and keeping track of animals for the Animal Census, welfare, and safety. Accurate animal counts help monitor population health, identify endangered species, and detect signs of disease or distress. Ensuring animal safety involves tracking movements to prevent poaching and human-wildlife conflicts. To address these needs, Codemonk worked with Kimzuka Solutions and the Forest Department of Karnataka to automate image classification, making wildlife research and conservation efforts more efficient.
Industry
Greentech
Time Frame
6 Months
Services we offered
AI/ML, Computer Vision, Product Engineering
Forest representatives faced the daunting task of meticulously sorting through hundreds of thousands of camera trap images. This manual process could take anywhere from six months to two years, significantly delaying census reporting and thereby conservation efforts. Timely actions against poaching or threats from carnivores to nearby small towns were hindered by this laborious process. Additionally, technical challenges arose from factors such as:
Similar Species
Differentiating between visually similar animals (e.g., wild dogs and jackals) proved difficult even for trained personnel.
Image Quality
Imperfect camera placement, false triggers, and nighttime captures produced images with inconsistent quality or clarity.
Data Imbalance
The frequency of certain animals in the dataset was significantly higher than others, creating potential bias in generic automation systems.
Codemonk's Approach
Codemonk approached this wildlife conservation challenge with a dual focus on technological advancement and user-centricity. We developed an AI-driven image classification solution tailored for the needs of Kimzuka Solutions and the Forest Department of Karnataka. We began by collaborating closely with Kimzuka Solutions to deeply understand their manual workflow, pain points, and goals. Recognizing the critical need for reliable data, we emphasized model accuracy through rigorous training, meticulous dataset preparation, and strategies to avoid bias. The solution was designed with the user in mind, delivered as an intuitive desktop application for easy use by Forest Department Staff. Understanding the variable nature of camera trap data, we built adaptability into the solution, allowing it to effectively handle diverse image quality, species, and complex real-world scenarios.
The Solution
Working alongside Kimzuka Solutions, Codemonk developed a robust AI-powered solution. The key components included:
Image Preprocessing
Images were enhanced with contrast adjustment, noise reduction, and color correction to ensure optimal quality for analysis.
Careful Dataset Curation
The training dataset was meticulously prepared to avoid bias and ensure the model's ability to recognize diverse animal species.
Sophisticated Animal Classifier
A cutting-edge model was trained to accurately identify animals, even when image quality or species similarity posed a challenge.
User-Friendly Application
The solution was delivered as an intuitive desktop application, empowering Kimzuka Solutions to efficiently analyze images and gain insights.
Impact and Outcome
The AI solution has revolutionized Kimzuka Solutions' operations. A staggering 180x reduction in image analysis time allows for the rapid progression of conservation initiatives. Now requiring minimal human oversight, resources can be optimally allocated elsewhere. The AI model's ability to surpass human accuracy in identifying animals from challenging images ensures reliable data at a faster pace. This enhanced efficiency and accuracy empower Forest Dept. of Karnataka to proactively monitor wildlife activity, swiftly responding to threats like poaching or potential risks posed by carnivores.
Reduction in Image Analysis Time
Reduction in Manual Effort