GroundingDINO
SmashBot
GroundingDINO | SmashBot | |
---|---|---|
5 | 6 | |
5,075 | 545 | |
8.3% | - | |
6.3 | 2.7 | |
9 days ago | 4 days ago | |
Python | Python | |
Apache License 2.0 | GNU General Public License v3.0 only |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
GroundingDINO
-
Autodistill: A new way to create CV models
Some of the foundation/base models include: * GroundedSAM (Segment Anything Model) * DETIC * GroundingDINO
-
Is there a way to do segmentation of a person's clothing?
While Segment Anything can detect objects based on text prompts, that's not its strong suite. To get best results, folks usually combine it with Grounding DINO, which is a great object detection model. You run Grounding DINO with text prompt "skirt", this gives you a bounding box that you pass to Segment Anything, which gives you a segmentation mask that you can then use for inpainting with SD.
-
Searching for Guidance on Developing an AI Bot for SSBU Training
Now, let's delve into the technological aspects of this project. The combination of Facebook's Segment Anything and Grounding Dino tools will automate annotations for image processing, which is key to this AI endeavor. I'm also intrigued by Mojo, a new programming language designed specifically for AI developers, which will soon be open-source.
-
[D] Object Detection Machine Learning
Right now we are trying out grouding dino on this but it is giving a lot of noise and detecting things that are not cracks.
- [D] Data Annotation Done by Machine Learning/AI?
SmashBot
- OpenAI Melee Bot?
-
Searching for Guidance on Developing an AI Bot for SSBU Training
There’s SmashBot in Melee- the bigger problem is probably trying to get it to learn how to play the entire roster and how to deal with matchups.
-
what's up with all the comparisons to NFTs??
More types of both will surely be developed in the future too, which is a big problem for assumptions about the limits "of AI" based only on our current models. We've seen this in other spaces too -- besides Deep Blue and AlphaGO turning people's expectations on their heads, smashbot for Super Smash Bros Melee was literally inspired by a bet that the game was too complex and required enough human "intuition" for an AI to ever compete with tournament-ranked players.
-
Valve bans 40k accounts after laying a trap for cheaters in Dota 2
There are bots for fighting games like the Rzr Infiltration bot for SFV and Smashbot for SSBM[1] which are good enough to take games off pro players. I'm not aware of it being abused at scale. Most people cheating in shooters or mobas aren't using full game-playing agents. They're using aimbots/skillshot scripts where you still need to move around and interact with stuff. Similarly, cheating in fighting games is typically done with button macros, so there's constant discussion on controller legality. However, button macros will only get you so far in fighting games, while an aimbot can get you close to the top of the ladder if undetected.
https://github.com/altf4/SmashBot
-
What artificial intelligence has been made for Smash?
There have been several attempts over the years for ssbm, like Philip the falcon ai and smashbot: https://github.com/altf4/SmashBot.
-
Is it possible to datamine/reverse engineer the SSBU ROM to see the source code the CPUs use? Feels like it might be really helpful in not-obvious ways
If you're interested in this stuff as a more academic question, have you seen the tasbot for melee, which is a genuine attempt at an optimal AI? The high level flowchart for that is pretty transparent: parry everything the opponent does, then whenever they're in enough lag for you to shine them, do a frame perfect waveshine 0-death. But there's more going on behind the scenes, and this is a bot which openly abuses it's nonhuman abilities to try to compete against top level players. It's open source, take a look! https://github.com/altf4/SmashBot
What are some alternatives?
segment-anything - The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example notebooks that show how to use the model.
libmelee - Open Python 3 API for making your own Smash Bros: Melee AI that works with Slippi Online
Detic - Code release for "Detecting Twenty-thousand Classes using Image-level Supervision".
OG-Snake-Game - The Original Snake Game. Maneuver a snake in its burrow and earn points while avoiding the snake itself and the walls of the snake burrow.
ultralytics - NEW - YOLOv8 🚀 in PyTorch > ONNX > OpenVINO > CoreML > TFLite
MultimediaVideo - Multimedia Video with Pygame and Opencv
CLIP - CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image
nintendo-games-ratings - Dataset and visualizations of Nintendo Games and ratings, scraped from metacritic.com
super-gradients - Easily train or fine-tune SOTA computer vision models with one open source training library. The home of Yolo-NAS.
crispy - Crispy is a machine-learning algorithm to make video-games montages efficiently. It uses a neural network to detect highlights in the video-game frames
LAVIS - LAVIS - A One-stop Library for Language-Vision Intelligence
OFA - Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework