Field Crop Health Analysis via Image Segmentation

Added Jun 2025 3 design docs

Assessing crop health across a field by eye is slow and inconsistent, and by the time stress or disease is visible during a casual walk-through, yield has often already been affected. Agronomists and agricultural researchers need a way to turn ordinary crop photographs into quantitative health measurements they can track over time. This project has the intern build a Streamlit application that segments crop imagery using two complementary approaches: classical computer vision methods, implementing Watershed and GrabCut with OpenCV, and a deep learning path using a U-Net style model built with Keras and TensorFlow. Once plants are isolated from soil and background, the app computes health indicators from the segmentation masks, such as vegetation coverage ratios and color-based stress signals derived with NumPy, and presents side-by-side visual comparisons of each technique's output so users can judge where classical methods suffice and where learned models win. By finishing the project, the intern develops a practical understanding of both traditional and neural segmentation pipelines, learns to derive domain-meaningful metrics from pixel masks rather than stopping at pretty pictures, and demonstrates the ability to package agricultural analytics into an interactive tool a non-programmer can operate.

Related projects

Crop Disease Classification Service Using Transfer Learning

Crop diseases are often identified too late for effective intervention because farmers and field agronomists rarely have quick access to plant pathology expertise. A photograph of a diseased leaf contains enough signal for accurate diagnosis, but turning that signal into a dependable, scalable service is a genuine engineering challenge for agricultural research groups. The intern builds a full diagnostic platform around transfer learning: pre-trained ResNet and MobileNet backbones are fine-tuned in PyTorch on crop disease imagery, with OpenCV and NumPy handling preprocessing and a data augmentation pipeline improving robustness to lighting and orientation. A Streamlit interface lets users upload leaf photos and receive ranked disease predictions alongside explainable AI visualizations that highlight the image regions driving each classification, while a FastAPI service exposes a batch inference endpoint for research workloads and pandas powers summary reporting. The entire system is containerized with Docker so it can be deployed consistently across lab and cloud environments. The project demonstrates end-to-end applied computer vision: adapting pre-trained networks to a specialist domain, communicating model reasoning to non-technical users, and packaging research code as a deployable, reproducible service.

agricultureeducationcomputer-visiondocker+9

View project

Medium

Wildlife Species Identification with Fine-Tuned ResNet Models

Conservation researchers collect enormous volumes of wildlife imagery from camera traps and field surveys, but manually identifying the species in every frame is a bottleneck that delays population studies and habitat decisions. Automated species classification lets small research teams process seasons of imagery in hours instead of months. The intern builds a Streamlit application centered on transfer learning: a ResNet model pre-trained on general imagery is fine-tuned in PyTorch to classify wildlife photos into multiple species. The project includes a data augmentation pipeline using OpenCV and NumPy to simulate the messy conditions of field photography, controls for staged fine-tuning of the network layers, and a prediction dashboard that reports ranked species probabilities for each uploaded image. An explanation view highlights the image regions most responsible for each prediction, helping ecologists confirm the model is keying on the animal rather than the background, while pandas tracks evaluation metrics across training runs. The intern comes away understanding how to adapt large pre-trained vision models to a specialized dataset, how augmentation and fine-tuning choices affect generalization, and how to present model evidence in a form that earns the trust of scientific users.

agricultureresearchcomputer-visionnumpy+5

View project

Easy

Farm Sensor Data Ingestion Pipeline with Kafka and HDFS

Modern farms generate continuous streams of readings for soil moisture, temperature, humidity, and equipment status, but the raw numbers are useless until they are reliably collected, stored, and summarized. Agritech companies solve this with streaming data platforms, and this project introduces that architecture at a scale a newcomer can build end to end. The intern writes Python producers that simulate realistic agricultural sensor streams and publish them to Kafka topics, then builds consumers that read the streams, validate and enrich the readings, and land them in HDFS for durable, distributed storage. On top of the stored data, Python analysis scripts compute aggregates such as average soil moisture per field, daily temperature ranges, and gaps that indicate sensor failures, producing summary files a farm manager could act on. The focus stays on the pipeline itself: partitioning topics sensibly, handling malformed messages, and organizing HDFS directories for easy downstream analysis. Through this big-data project the intern gains a concrete understanding of publish-subscribe streaming, distributed file storage, and pipeline design, which are foundational skills for any data engineering role.

agriculturetechnologybig-datahdfs+2

View project