LTTM Laboratory

18/06/2021

is starting tomorrow (Saturday, June 19th)! This week ahead is going to be one of the most exciting in the Computer Vision field.
Make sure you don't miss out!

We will present 3 works (in chronological order):

1) Sunday, June 19th at Workshop on Autonomous Driving
"Latent Space Regularization for Unsupervised Domain Adaptation in Semantic Segmentation", where we improve feature-level disentanglement for better recognition when translating between different visual domains.
Francesco Barbato, Marco Toldo, Umberto Michieli and Pietro Zanuttigh
https://lnkd.in/d6_hm8R

2) Sunday, June 19th at Workshop on Resposible Computer Vision
"Are All Users Treated Fairly in Federated Learning Systems?", where we analyse fairness of common federated learning frameworks.
Umberto Michieli and Mete Ozay. Researched during Umberto's internship at Samsung Research UK
https://lnkd.in/dwaKRrr

3) Monday, June 20th at the first session of the main conference
"Continual Semantic Segmentation via Repulsion-Attraction of Sparse and Disentangled Latent Representations", where we leverage continual semantic segmentation by constraining the latent space to preserve representations of old tasks, thus mitigating catastrophic forgetting.
Umberto Michieli and iPietro Zanuttigh
https://lnkd.in/dxNnW3S

we hope to meet you in the virtual platform!

Huge thanks to all the authors for the wonderful teamwork.

17/06/2021

LTTM participated to the GTTI seminar on 'Deep signal processing for a safer world' with two projects presented by Daniele and Francesco!

16/06/2021

Deep learning has many applications and if you want to know more, don't miss tomorrow's meeting on deep learning for forensic and security https://unipd.zoom.us/j/87930477008?pwd=bTNNQ2xEUldDdWVEMVZLQys2bGJNUT09 ! The program is in the flyer!

11/06/2021

Here's the video of our presentation! Enjoy!

Here's the GTTI preview of ICASSP 2021 video

09/06/2021

This year ICASSP 2021 is avatarized! The small digital mini-me is wandering in the conference hall looking for the poster room 🙂.

See you on session IFS-4 Surveillance, Biometrics and Security tonight @ 10:30pm CEST to discuss our work

"LOOKING THROUGH WALLS: INFERRING SCENES FROM VIDEO-SURVEILLANCE ENCRYPTED TRAFFIC"

03/06/2021

Is you video surveillance system completely safe? Probably, not.

A man is walking along a corridor, monitored by a set of CCTV cameras spread all over the building. Cameras are transmitting the encrypted video sequences to a central control room so that, even if a possible malicious user sniffs the stream, he/she can not decode and visualize the information … are we sure?

In the paper “Looking Through Walls: Inferring Scenes from Video-Surveillance Encrypted Traffic”, we are investigating how much information can be estimated from an encrypted video stream by analyzing packet lengths. The work was carried on within a cross-department collaboration involving Daniele Mari, Samuele Giuliano Piazzetta, Sara Bordin, Luca Pajola, Sebastiano Verde, Simone Milani, Mauro Conti.

Check out the full paper or join us at ICASSP 2021 next week.

01/06/2021

This month we are going to present our recent paper at CVPR 2021.

In our work, we teach machines to automatically improve the learned knowledge to recognize and segment new classes. Our main intuition is that if features are representative of their class and well-separated, then it’s easier to preserve high accuracy on them whilst accommodating future classes. We achieve this by disentangling the feature space: we constrain features of the same classes to be tightly clustered, to be spaced apart from features of other classes and to be sparse. In this way, we preserve geometrical space, since few channels are active, and expressiveness, since features are divided into well-separated clusters. Then, it is easier to accommodate the latent representations of future classes.

Our technology will be helpful for agents deployed in the real world, which continuously need to adapt their discriminative knowledge.

Paper webpage: https://lttm.dei.unipd.it/paper_data/SDR/

28/05/2021

If you are interested in Point Cloud coding don't miss the webinar by Fernando Pereira on June 2nd at 12:30! You can subscribe at bit.ly/3bD3NyZ
See you there!

19/05/2021

There are only a couple of days left to register to the Winter Sports Hackathon that will take place on June 8 and 9! If you want to know more, check out the website!

The hackathon is open to all students and fans of data and Winter Sports, the teams will have 36 hours to extract from sensors data and video sequences useful information, metrics, and comparisons in order to improve/innovate fan engagement systems.

18/05/2021

Today we showcased the project 'Mixed Reality Dance' in the virtual visit to the labs within the Master Degree in ICT for Internet and Multimedia. Thanks to Elena and Daniele, Elvis danced samba and belly dance in our lab!!

13/10/2016

LTTM@eccv2016 !!