Rafael Padilla
Machine Learning Engineer at Laboratório de Sinais, Multimídia e Telecomunicações - SMT
Summary
🤩
RockstarRafael Padilla is a machine learning engineer from Brazil with eight years of hands-on experience specializing in computer vision and anomaly detection. He contributes to prominent open-source projects—most notably improving object detection functionality within Hugging Face Transformers—and maintains widely used repositories for object detection metrics and evaluation. Rafael’s strengths lie in designing robust bounding-box representations, IoU calculations, and format converters (COCO, PASCAL, YOLO, CVAT), enabling reliable evaluation pipelines across datasets. He blends back-end engineering and data-science rigor to improve precision-recall and mAP computations, and is comfortable making production-focused changes to popular ML libraries. An inquisitive practitioner, he often surfaces practical fixes like deprecations, threshold tuning, and format edge cases that reduce friction for applied vision teams.
8 years of coding experience
Github Skills (16)
transformers10
data-structures10
object-detection10
model-building10
computer-vision10
pytorch10
machine-learning10
bounding-box10
python10
modeling10
data-structure10
metric10
image-processing10
model-driven-development10
model-driven10
Programming languages (5)
Github contributions (5)
rafaelpadilla/review_object_detection_metrics
Nov 2020 - Aug 2022
Object Detection Metrics. 14 object detection metrics: mean Average Precision (mAP), Average Recall (AR), Spatio-Temporal Tube Average Precision (STT-AP). This project supports different bounding box formats as in COCO, PASCAL, Imagenet, etc.
Role in this project:
Back-end Developer & Data Scientist
Back-end Developer & Data ScientistContributions:17 reviews, 170 commits, 11 PRs in 1 year 9 months
Contributions summary:Rafael primarily contributed to the development of core functionalities within the object detection metrics project. They focused on implementing the `BoundingBox` class, a foundational component of the project, and created methods for absolute and relative bounding box formats, including the implementation of the IOU calculation methods. The commits also show the user's involvement in creating and adapting converters for diverse annotation formats such as COCO, CVAT, and YOLO, demonstrating expertise in handling different data structures within the object detection domain.
precision-recallboundingobject-detection-metricstubebounding-box
rafaelpadilla/Object-Detection-Metrics
May 2018 - Aug 2022
Most popular metrics used to evaluate object detection algorithms.
Role in this project:Back-end Developer
Back-end DeveloperContributions:2 releases, 3 reviews, 108 commits in 4 years 3 months
Contributions summary:Rafael primarily contributed to the `Evaluator.py` file, adding and modifying functionalities related to object detection metrics. These changes included implementing the 11-point interpolation method, correcting comparisons, adding comments, and fixing height when the format is XYX2Y2. The modifications centered around the precision-recall curve, improving its presentation and refining the calculation of average precision. The user also made adjustments to the sample code files.
pascal-vocevaluateprecision-recalldeep-learningevaluation
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.