Gustavo A. Salazar-Gomez

I'm a PhD student at Universite Grenoble Alpes, under the supervison of Anne Spalanzani and Christian Laugier. I work at INRIA's Chroma lab, where I focus my research in AI-driven algorithms for safe motion planning and driving decision-making in the context of autonomous driving.

Also, I completed a MSc in Mobile, Autonomous and Robotic Systems at Grenoble INP in 2022, before that, a Postgraduate Diploma in Artificial Intelligence in 2021 and BEng as Mechatronics Engineer in 2018 from Universidad Autonoma de Occidente (UAO).

Email: gustavo [hyphen] andres [dot] salazar [hyphen] gomez [at] univ [hyphen] grenoble [hyphen] alpes [dot] fr

Github  /  Google Scholar  /  Linkedin  /  Email  /  CV

• [June 2024] Presented the paper "TLCFuse: Temporal Multi-Modality Fusion Towards Occlusion-Aware Semantic Segmentation" at 2024 IEEE Intelligent Vehicles Symposium (IV) in Jeju, South Korea.
• [January 2024] Started my PhD studies at Universite Grenoble Alpes and INRIA, under the supervison of Anne Spalanzani and Christian Laugier.
• [December 2022] Presented the paper "TransFuseGrid: Transformer-based Lidar-RGB fusion for semantic grid prediction" at 2022 17th International Conference on Control, Automation, Robotics and Vision (ICARCV).
• [October 2022] Started a new position as Research Engineer at INRIA's Chroma lab, where I focused my research in multi-modal sensor fusion for autonmous driving.
• [September 2022] Defended my Master thesis "Transformer-based Lidar-RGB Fusion for Semantic Grid Prediction in Autonomous Vehicles" performed at INRIA's Chroma lab.
• [March 2022] Started an internship at INRIA's Chroma lab, exploring sensor fusion techniques for robust situational awareness for Autonomous Vehicles.
• [Sept 2021] Started a MSc in Mobile, Autonomous and Robotic Systems at Grenoble INP.
• [June 2021] Finished Postgraduate Diploma in Artificial Intelligence at Universidad Autonoma de Occidente.
• [June 2021] The abstract High-level camera-LiDAR fusion for 3D object detection with Machine Learning was presented at the LatinX in AI Research Workshop at CVPR 2021 as a poster presentation.
• [August 2020] Started a Postgraduate Diploma in Artificial Intelligence at Universidad Autonoma de Occidente, the degree will end in June 2021.
• [January 2019] The article "Application of Transfer Learning for Object Recognition Using Convolutional Neural Networks" was selected to be published as an extended version in Applications of Computational Intelligence book part of the Springer Communications in Computer and Information Science book series (CCIS, volume 833).
• [May 2018] The paper "Application of Transfer Learning for Object Recognition Using Convolutional Neural Networks" was published in the conference IEEE Colombian Conference on Applications in Computational Intelligence (ColCACI).

In this paper, we emphasize the crucial role of temporal cues in reinforcing resilience against occlusions in the bird’s eye view (BEV) semantic grid segmentation task. We proposed a novel architecture that enables the processing of temporal multi-step inputs, where the input at each time step comprises the spatial information encoded from fusing LiDAR and camera sensor readings.

In this work, we presented an approach for semantic grid prediction that uses a transformer architecture to fuse Lidar sensor data with RGB images from multiple cameras to predict semantic grids of the ego vehicle's environment.

3D object detection of vehicles in the NuScenes dataset using classic Machine learning such as DBSCAN and SVMs.

A transfer learning technique is used to create a computational tool that recognizes the objects of the automatics laboratory of the Universidad Autónoma de Occidente in real time. The Inception-V3 is used as a pre-trained CNN.

Style-transfer implementarion based on the paper A neural algorithm of artistic style using VGG-19.

Convolutional Neural Network Models LeNet5 and VGG16 with Tensorflow for MNIST dataset, trained with GPU on Colab.

Machine Learning projects from Supervised methods like KNN, Decision Trees and ANN, and Unsupervised methods like DBSCAN and PCA, finally feature engineering to improve models performance.

Generative Adversarial Networks(GAN) used to generate cars images from CIFAR10, trained with GPU on Colab.

Face recognition using LBP features, the system extract the LBP features from the face in the image and a SVM is trained in order to classify each face. There could be multiple faces in the scene.

Sudoku recognizer using OpenCV to process the image, and a KNN as ML model to classify each digit found, finally the result is displayed in a grid where the numbers are in the same position as the image.

• Introduction to Satellite Communications by Institut Mines-Télécom on Coursera. Certificate earned on January 2, 2020. [Credential]