In this section, I have compiled the main projects that I have completed since the last years of my degree.
Personal projects
Fish segmentation
Pytorch project for fish segmentation (2021)
Developed a convolutional neural network for semantic segmentation of fish in images, utilizing the PyTorch library for training and inference.
- Repository: https://github.com/isaacperez/fish-segmentation
Tensorflow 1.2 model to Raspberry Pi C code
Transforming a Tensorflow 1.2 model to Raspberry Pi C code (2018)
Created a streamlined method to convert a TensorFlow model into C code suitable for running on a Raspberry Pi, enhancing accessibility and deployment.
- Repository: https://github.com/isaacperez/tf12raspberry
Embedded solution
NVIDIA Jetson TX2 + IntelRealSense Depth camera + ROS + OpenCV + VisionWorks (2018)
Developed an embedded solution integrating NVIDIA Jetson TX2 with Intel RealSense cameras using ROS, OpenCV, and VisionWorks, for advanced image processing applications.
- Repository: https://github.com/isaacperez/embedded-solution
Professional projects
Tool for pathology recognition
Tool for pathology recognition in capsule endoscopic images using deep learning (2021) Collaborated with Juan Ramón Jiménez Hospital to develop a deep learning tool to assist specialists in detecting pathologies in capsule endoscopic images.
Automated biodiversity monitoring system based on deep learning and citizen science
Automated biodiversity monitoring system in the Iberian Peninsula (2020)
Developed a deep learning system for classifying camera trap images and launched a citizen science project on Zooniverse, achieving significant engagement.
Zoonvierse project: https://www.zooniverse.org/projects/aicensusuhu/iberian-camera-trap-project
Press release: https://theconversation.com/la-inteligencia-artificial-nos-ayuda-a-estudiar-la-fauna-de-donana-para-mejorar-su-conservacion-162815+
EyeVIEW
System for the detection of printing errors on metal foils (2018)
Designed and implemented a computer vision solution for inspecting prints on metal foils, developed entirely in C++.
Strawberry instance segmentation
Strawberry instance segmentation at Agrobot (2017)
Developed an advanced strawberry instance segmentation algorithm for Agrobot’s harvesting robot. Featured by NVIDIA in the I AM AI Docuseries.
- NVIDIA programme: https://www.youtube.com/watch?v=bXQg_M7_6_E
Academic projects
Blood vessel segmentation
A new deep learning method for blood vessel segmentation in retinal images based on convolutional kernels and modified U-Net model (2021)
Modified the U-Net network and introduced new training techniques to achieve state-of-the-art performance in blood vessel segmentation.
Strawberry instance segmentation - Round 2
A new deep-learning strawberry instance segmentation methodology based on a fully convolutional neural network (2021)
Introduced a new instance segmentation paradigm using fully convolutional neural networks, outperforming Mask R-CNN in speed and accuracy.
DEEP LEARNING
DEEP LEARNING: Fundamentals, Theory and Application (2021)
Authored a comprehensive book on deep learning, covering neural networks, recurrent neural networks, and convolutional neural networks with detailed illustrations. Available for free.
Strawberry instance segmentation - Round 1
A fast and accurate deep learning method for strawberry instance segmentation (2020)
Enhanced Mask R-CNN to double its fps without losing accuracy, optimizing the model for real-world performance-limited environments.
StrawDI
The Strawberry Digital Images Data Set (2020)
Created the first public dataset of strawberry images for instance segmentation, containing 8000 images with manually labeled ground truth data.
- Download: https://strawdi.github.io/
Student projects
Master’s Degree Final Project
Strawberries instance segmentation for an automated strawberry harvester (2020)
Implemented Mask R-CNN from scratch to achieve individual segmentation of strawberries, using images collected from plantations for model training.
Bachelor’s Degree Final Project
Detection, segmentation and classification of strawberries for an automated strawberry harvester (2017)
Designed a system using three convolutional neural networks to detect, segment, and classify strawberries based on ripeness. Included component selection and computer setup for model training.
The Mario AI Benchmark
Intelligent agent to play The Mario AI competition (2017)
Developed a JAVA-based agent trained with reinforcement learning using the Q-learning algorithm to play Super Mario for Mario AI competition.
Rule-based autonomous driver compiler for Torcs
Rules-based system for playing Torcs with a compiler that generates the player (2016)
Built a compiler that generates an autonomous driver for TORCS from a set of rules using fuzzy logic, developed in JAVA.