Data Science by ODS.ai 🦜
45.1K subscribers
754 photos
84 videos
7 files
1.83K links
First Telegram Data Science channel. Covering all technical and popular staff about anything related to Data Science: AI, Big Data, Machine Learning, Statistics, general Math and the applications of former. To reach editors contact: @malev
Download Telegram
​​Google announced the updated YouTube-8M dataset

Updated set now includes a subset with verified 5-s segment level labels, along with the 3rd Large-Scale Video Understanding Challenge and Workshop at #ICCV19.

Link: https://ai.googleblog.com/2019/06/announcing-youtube-8m-segments-dataset.html

#Google #YouTube #CV #DL #Video #dataset
​​Simultaneous food and facial recognition at a Foxconn factory canteen, Shenzhen China

#video #foodlearning #facerecogniction #dl #cv #foxconn
Deep Fake Challenge by Facebook team

#Facebook launches a competition to fight deep fakes. Unfortunately, results of this competition will be obviously used to create better fakes, to the cheers of the people, wishing to watch the Matrix with Bruce Lee or more questionable deep fake applications.

Link: https://ai.facebook.com/blog/deepfake-detection-challenge/

#deepfake #video #cv #dl
Castle in the Sky

Dynamic Sky Replacement and Harmonization in Videos

Fascinating and ready to be applied for work. (With colab notebook)
The authors proposed a method to replace the sky in the video that works well in high resolution. The results are very impressive. The method runs in real-time and produces video almost without glitches and artifacts. Also, can generate for example lightning and glow on target video.
The pipeline is quite complicated and contains several tasks:
– A sky matting network to segmentation sky on video frames
– A motion estimator for sky objects
– A skybox for blending where sky and other environments on video are relighting and recoloring.
Authors say their work, in a nutshell, proposes a new framework for sky augmentation in outdoor videos. The solution is purely vision-based and it can be applied to both online and offline scenarios.
But let's take a closer look.

A sky matting module is a ResNet-like encoder and several layers upsampling decoder to solve sky pixel-wise segmentation tasks followed by a refinement stage with guided image filtering.
A motion estimator directly estimates the motion of the objects in the sky. The motion patterns are modeled by an affine matrix and optical flow.
The sky image blending module is a decoder that models a linear combination of target sky matte and aligned sky template.

Overall, the network architecture is ResNet-50 as encoder and decoder with coordConv upsampling layers with skip connections and implemented in Pytorch,

The result is presented in a very cool video https://youtu.be/zal9Ues0aOQ


site: https://jiupinjia.github.io/skyar/
paper: https://arxiv.org/abs/2010.11800
github: https://github.com/jiupinjia/SkyAR


#sky #CV #video #cool #resnet
πŸ‘1
Forwarded from Machinelearning
⚑️ HunyuanCustom: консистСнтная видСогСнСрация c ΠΈΠ½ΠΏΠ΅ΠΉΠ½Ρ‚ΠΎΠΌ ΠΈ липсинком.

Tencent выпустила HunyuanCustom, Ρ„Ρ€Π΅ΠΉΠΌΠ²ΠΎΡ€ΠΊ, ΠΊΠΎΡ‚ΠΎΡ€Ρ‹ΠΉ Π½Π΅ Ρ‚ΠΎΠ»ΡŒΠΊΠΎ Π³Π΅Π½Π΅Ρ€ΠΈΡ€ΡƒΠ΅Ρ‚ Π²ΠΈΠ΄Π΅ΠΎ ΠΏΠΎ Π·Π°Π΄Π°Π½Π½Ρ‹ΠΌ условиям, Π½ΠΎ ΠΈ ΡƒΠΌΠ΅Π΅Ρ‚ ΡΠΎΡ…Ρ€Π°Π½ΡΡ‚ΡŒ ΠΊΠΎΠ½ΡΠΈΡΡ‚Π΅Π½Ρ‚Π½ΠΎΡΡ‚ΡŒ ΡΡƒΠ±ΡŠΠ΅ΠΊΡ‚ΠΎΠ², Π±ΡƒΠ΄ΡŒ Ρ‚ΠΎ Ρ‡Π΅Π»ΠΎΠ²Π΅ΠΊ, ΠΆΠΈΠ²ΠΎΡ‚Π½ΠΎΠ΅ ΠΈΠ»ΠΈ ΠΏΡ€Π΅Π΄ΠΌΠ΅Ρ‚. МодСль справляСтся Π΄Π°ΠΆΠ΅ с ΠΌΡƒΠ»ΡŒΡ‚ΠΈΡΡƒΠ±ΡŠΠ΅ΠΊΡ‚Π½Ρ‹ΠΌΠΈ сцСнами: Π² Π΄Π΅ΠΌΠΎ-Ρ€ΠΎΠ»ΠΈΠΊΠ°Ρ… люди СстСствСнно Π²Π·Π°ΠΈΠΌΠΎΠ΄Π΅ΠΉΡΡ‚Π²ΡƒΡŽΡ‚ с ΠΏΡ€Π΅Π΄ΠΌΠ΅Ρ‚Π°ΠΌΠΈ, Π° тСкст Π½Π° ΡƒΠΏΠ°ΠΊΠΎΠ²ΠΊΠ°Ρ… Π½Π΅ ΠΏΠ»Ρ‹Π²Π΅Ρ‚ ΠΌΠ΅ΠΆΠ΄Ρƒ ΠΊΠ°Π΄Ρ€Π°ΠΌΠΈ.

Π’ основС ΠΌΠΎΠ΄Π΅Π»ΠΈ Π»Π΅ΠΆΠΈΡ‚ ΡƒΠ»ΡƒΡ‡ΡˆΠ΅Π½Π½Ρ‹ΠΉ ΠΌΠ΅Ρ…Π°Π½ΠΈΠ·ΠΌ слияния тСкста ΠΈ ΠΈΠ·ΠΎΠ±Ρ€Π°ΠΆΠ΅Π½ΠΈΠΉ Ρ‡Π΅Ρ€Π΅Π· LLaVA. НапримСр, Ссли Π²Ρ‹ Π·Π°Π³Ρ€ΡƒΠΆΠ°Π΅Ρ‚Π΅ Ρ„ΠΎΡ‚ΠΎ ΠΆΠ΅Π½Ρ‰ΠΈΠ½Ρ‹ Π² ΠΏΠ»Π°Ρ‚ΡŒΠ΅ ΠΈ тСкст Β«Ρ‚Π°Π½Ρ†ΡƒΠ΅Ρ‚ ΠΏΠΎΠ΄ Π΄ΠΎΠΆΠ΄Π΅ΠΌΒ», систСма Π°Π½Π°Π»ΠΈΠ·ΠΈΡ€ΡƒΠ΅Ρ‚ ΠΎΠ±Π° ΠΈΠ½ΠΏΡƒΡ‚Π°, связывая описаниС с Π²ΠΈΠ·ΡƒΠ°Π»ΡŒΠ½Ρ‹ΠΌΠΈ дСталями.

Но Π³Π»Π°Π²Π½ΠΎΠ΅ - это ΠΌΠΎΠ΄ΡƒΠ»ΡŒ Π²Ρ€Π΅ΠΌΠ΅Π½Π½ΠΎΠΉ ΠΊΠΎΠ½ΠΊΠ°Ρ‚Π΅Π½Π°Ρ†ΠΈΠΈ: ΠΎΠ½ «растягиваСт» особСнности изобраТСния вдоль Π²Ρ€Π΅ΠΌΠ΅Π½Π½ΠΎΠΉ оси Π²ΠΈΠ΄Π΅ΠΎ, ΠΈΡΠΏΠΎΠ»ΡŒΠ·ΡƒΡ 3D-VAE. Π­Ρ‚ΠΎ ΠΏΠΎΠΌΠΎΠ³Π°Π΅Ρ‚ ΠΈΠ·Π±Π΅ΠΆΠ°Ρ‚ΡŒ Β«ΠΏΡ€Ρ‹Π³Π°ΡŽΡ‰ΠΈΡ…Β» Π»ΠΈΡ† ΠΈΠ»ΠΈ Π²Π½Π΅Π·Π°ΠΏΠ½Ρ‹Ρ… ΠΈΠ·ΠΌΠ΅Π½Π΅Π½ΠΈΠΉ Ρ„ΠΎΠ½Π°, ΠΏΡ€ΠΎΠ±Π»Π΅ΠΌΡ‹, которая Ρ…Π°Ρ€Π°ΠΊΡ‚Π΅Ρ€Π½Π° Π΄Π°ΠΆΠ΅ для Ρ‚ΠΎΠΏΠΎΠ²Ρ‹Ρ… ΠΌΠΎΠ΄Π΅Π»Π΅ΠΉ Π²ΠΈΠ΄Π΅ΠΎΠ³Π΅Π½Π΅Ρ€Π°Ρ†ΠΈΠΈ.

Tencent ΠΏΠ΅Ρ€Π΅Ρ€Π°Π±ΠΎΡ‚Π°Π»ΠΈ ΠΈ ΠΏΠ°ΠΉΠΏΠ»Π°ΠΉΠ½ Π°ΡƒΠ΄ΠΈΠΎ. Для синхронизации Π·Π²ΡƒΠΊΠ° с двиТСниями Π³ΡƒΠ± ΠΈΠ»ΠΈ дСйствиями Π² ΠΊΠ°Π΄Ρ€Π΅ HunyuanCustom ΠΈΡΠΏΠΎΠ»ΡŒΠ·ΡƒΠ΅Ρ‚ AudioNet, ΠΌΠΎΠ΄ΡƒΠ»ΡŒ, ΠΊΠΎΡ‚ΠΎΡ€Ρ‹ΠΉ Π²Ρ‹Ρ€Π°Π²Π½ΠΈΠ²Π°Π΅Ρ‚ Π°ΡƒΠ΄ΠΈΠΎ- ΠΈ Π²ΠΈΠ΄Π΅ΠΎΡ„ΠΈΡ‡ΠΈ Ρ‡Π΅Ρ€Π΅Π· пространствСнноС кросс-Π²Π½ΠΈΠΌΠ°Π½ΠΈΠ΅.

Π€Ρ€Π΅ΠΉΠΌΠ²ΠΎΡ€ΠΊ ΠΏΠΎΠ΄Π΄Π΅Ρ€ΠΆΠΈΠ²Π°Π΅Ρ‚ Π²ΠΎΠ·ΠΌΠΎΠΆΠ½ΠΎΡΡ‚ΡŒ Π·Π°ΠΌΠ΅Π½Ρ‹ ΠΎΠ±ΡŠΠ΅ΠΊΡ‚Π° Π² Π³ΠΎΡ‚ΠΎΠ²ΠΎΠΌ Ρ€ΠΎΠ»ΠΈΠΊΠ΅ (скаТСм, ΠΏΠΎΠ΄ΡΡ‚Π°Π²ΠΈΡ‚ΡŒ Π½ΠΎΠ²ΡƒΡŽ модСль кроссовок Π² Ρ€Π΅ΠΊΠ»Π°ΠΌΡƒ), модСль сТимаСт исходноС Π²ΠΈΠ΄Π΅ΠΎ Π² Π»Π°Ρ‚Π΅Π½Ρ‚Π½ΠΎΠ΅ пространство, Π²Ρ‹Ρ€Π°Π²Π½ΠΈΠ²Π°Π΅Ρ‚ Π΅Π³ΠΎ с ΡˆΡƒΠΌΠ½Ρ‹ΠΌΠΈ Π΄Π°Π½Π½Ρ‹ΠΌΠΈ ΠΈ встраиваСт измСнСния Π±Π΅Π· Π°Ρ€Ρ‚Π΅Ρ„Π°ΠΊΡ‚ΠΎΠ² Π½Π° Π³Ρ€Π°Π½ΠΈΡ†Π°Ρ….

Π­ΠΊΡΠΏΠ΅Ρ€ΠΈΠΌΠ΅Π½Ρ‚Π°Π»ΡŒΠ½Ρ‹Π΅ тСсты ΠΏΠΎΠΊΠ°Π·Π°Π»ΠΈ, Ρ‡Ρ‚ΠΎ HunyuanCustom ΠΎΠ±Ρ…ΠΎΠ΄ΠΈΡ‚ ΠΊΠΎΠ½ΠΊΡƒΡ€Π΅Π½Ρ‚ΠΎΠ² ΠΏΠΎ ΠΊΠ»ΡŽΡ‡Π΅Π²Ρ‹ΠΌ ΠΌΠ΅Ρ‚Ρ€ΠΈΠΊΠ°ΠΌ. НапримСр, Face-Sim (сохранСниС идСнтичности Π»ΠΈΡ†Π°) Ρƒ Tencent β€” 0.627 ΠΏΡ€ΠΎΡ‚ΠΈΠ² 0.526 Ρƒ Hailuo, Π° с Keling, Vidu, Pika ΠΈ Skyreels Ρ€Π°Π·Ρ€Ρ‹Π² Π΅Ρ‰Π΅ большС.

⚠️ Для Ρ€Π°Π±ΠΎΡ‚Ρ‹ модСль Ρ‚Ρ€Π΅Π±ΡƒΠ΅Ρ‚ ΠΌΠΈΠ½ΠΈΠΌΡƒΠΌ 24 Π“Π‘ видСопамяти для Ρ€ΠΎΠ»ΠΈΠΊΠΎΠ² 720p, Π½ΠΎ Ρ‡Ρ‚ΠΎΠ±Ρ‹ Ρ€Π°ΡΠΊΡ€Ρ‹Ρ‚ΡŒ всС возмоТности, Ρ€Π°Π·Ρ€Π°Π±ΠΎΡ‚Ρ‡ΠΈΠΊΠΈ Ρ€Π΅ΠΊΠΎΠΌΠ΅Π½Π΄ΡƒΡŽΡ‚ 80 Π“Π‘ VRAM.

Код ΠΈ Ρ‡Π΅ΠΊΠΏΠΎΠΈΠ½Ρ‚Ρ‹ ΡƒΠΆΠ΅ доступны Π² ΠΎΡ‚ΠΊΡ€Ρ‹Ρ‚ΠΎΠΌ доступС, Π° Π² Ρ€Π΅ΠΏΠΎΠ·ΠΈΡ‚ΠΎΡ€ΠΈΠΈ Π΅ΡΡ‚ΡŒ ΠΏΡ€ΠΈΠΌΠ΅Ρ€Ρ‹ запуска ΠΊΠ°ΠΊ Π½Π° Π½Π΅ΡΠΊΠΎΠ»ΡŒΠΊΠΈΡ… GPU, Ρ‚Π°ΠΊ ΠΈ Π² экономном Ρ€Π΅ΠΆΠΈΠΌΠ΅ для ΠΏΠΎΡ‚Ρ€Π΅Π±ΠΈΡ‚Π΅Π»ΡŒΡΠΊΠΈΡ… Π²ΠΈΠ΄Π΅ΠΎΠΊΠ°Ρ€Ρ‚.


πŸ“ŒΠ›ΠΈΡ†Π΅Π½Π·ΠΈΡ€ΠΎΠ²Π°Π½ΠΈΠ΅ ΠΊΠΎΠ΄Π° : Tencent Hunyuan Community License.


πŸŸ‘Π‘Ρ‚Ρ€Π°Π½ΠΈΡ†Π° ΠΏΡ€ΠΎΠ΅ΠΊΡ‚Π°
🟑МодСль
🟑Arxiv
πŸ–₯GitHub


@ai_machinelearning_big_data

#AI #ML #Video #HunyuanCustom #Tencent
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
πŸ‘8πŸ”₯4πŸ₯°2