Try out deep learning models online on Colab with a single click.

Image Super-Resolution

• Real Cascade U-Nets for Anime Image Super Resolution with bilibili/ailab.
• Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution with JingyunLiang/MANet.
• Image Super-Resolution via Iterative Refinement with Janspiry/Image-Super-Resolution-via-Iterative-Refinement.
• SwinIR: Image Restoration Using Swin Transformer with JingyunLiang/SwinIR.
• Enhanced Super-Resolution Generative Adversarial Networks (ESRGAN). A combination of xinntao/ESRGAN and ata4/esrgan-launcher.
  • My colab fork is located in styler00dollar/Colab-ESRGAN.
  • Alternative with efonte/ESRGAN.
  • Train ESRGAN with forked xinntao/BasicSR (victorca25/traiNNer) and my fork styler00dollar/Colab-traiNNer.
• Real-ESRGAN aims at developing Practical Algorithms for General Image Restoration with xinntao/Real-ESRGAN.
  • My own minimalistic version with styler00dollar/Colab-Real-ESRGAN:
  • Modified with different scales and pixel unshuffle with Real-ESRGAN-colab.
• Deep Unfolding Network for Image Super-Resolution (USRNet). The original repositories are cszn/USRNet and cszn/KAIR. My colab fork is located in styler00dollar/Colab-USRNet.
• Image Super Resolution with idealo/image-super-resolution.
• PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models. or
• SPSR: Structure-Preserving Super Resolution with Gradient Guidance with Maclory/SPSR. A Colab based on BlueAmulet/SPSR is located in my fork styler00dollar/Colab-SPSR.

Video Interpolation

• Asymmetric Bilateral Motion Estimation for Video Frame Interpolation with JunHeum/ABME. My fork is located in styler00dollar/Colab-ABME.
• Anime interpolation with lisiyao21/AnimeInterp. My fork is located in styler00dollar/Colab-AnimeInterp.
• RIFE: Real-Time Intermediate Flow Estimation for Video Frame Interpolation with hzwer/arXiv2020-RIFE. My fork with FFMPEG is located in styler00dollar/Colab-RIFE.
  • Alternative with HeylonNHP/RIFE-Colab.
• Depth-Aware Video Frame Interpolation (DAIN) using baowenbo/DAIN. My fork is located in styler00dollar/Colab-DAIN.
  • DAIN NCNN with nihui/dain-ncnn-vulkan. My Colab fork is located in styler00dollar/Colab-dain-ncnn-vulkan.
• Channel Attention Is All You Need for Video Frame Interpolation with myungsub/CAIN. I modified Hubert482/cainapp with my fork styler00dollar/Colab-CAIN.
  • CAIN NCNN (Channel Attention Is All You Need for Video Frame Interpolation) with nihui/cain-ncnn-vulkan. My fork is located in styler00dollar/Colab-cain-ncnn-vulkan.
• High Quality Estimation of Multiple Intermediate Frames for Video Interpolation with avinashpaliwal/Super-SloMo. My colab fork is locaced in styler00dollar/Colab-Super-SloMo. My version: The original:
• Featureflow can be found in CM-BF/FeatureFlow. The colab was made by Mr. Anon.
• Video Frame Interpolation via Residue Refinement with HopLee6/RRIN. My fork is located in styler00dollar/Colab-RRIN.

Inpainting

• Resolution-robust Large Mask Inpainting with Fourier Convolutions with saic-mdal/lama.
  • Official Colab:
  • My Colab styler00dollar/Colab-lama:
• Learning a Sketch Tensor Space for Image Inpainting of Man-made Scenes with ewrfcas/MST_inpainting. My fork is located in styler00dollar/Colab-MST.
• Generative Image Inpainting with Auxiliary Contextual Reconstruction with zengxianyu/crfill. My fork is located in styler00dollar/Colab-crfill.
• Large Scale Image Completion via Co-Modulated Generative Adversarial Networks with zsyzzsoft/co-mod-gan. My fork is located in styler00dollar/Colab-co-mod-gan.
  • There is a pytorch version with zengxianyu/co-mod-gan-pytorch. My fork is located in styler00dollar/Colab-co-mod-gan-pytorch.
• Image Completion Transformer with raywzy/ICT. My fork is located in styler00dollar/Colab-ICT.
• DFNet: Deep Fusion Network for Image completion with hughplay/DFNet. My fork with Yukariin/DFNet is located in styler00dollar/Colab-DFNet.
• open-mmlab/mmediting is an open source image and video editing toolbox based on PyTorch. My fork is located in styler00dollar/Colab-mmediting.
• EdgeConnect: Generative Image Inpainting with Adversarial Edge Learning with knazeri/edge-connect. My fork is located in styler00dollar/Colab-edge-connect.
• Rethinking Inpainting with KumapowerLIU/Rethinking-Inpainting-MEDFE. My fork is located in styler00dollar/Colab-MEDFE. [WARNING: NEEDS MATLAB]
• Region Normalization for Image Inpainting with geekyutao/RN. My fork is located in styler00dollar/Colab-RN.
• Coherent Semantic Attention Image Inpainting with Yukariin/CSA_pytorch. My fork is located in styler00dollar/Colab-CSA-pytorch.
• Pluralistic Image Completion with lyndonzheng/Pluralistic-Inpainting. My fork is located in styler00dollar/Colab-Pluralistic-Inpainting.
• SayedNadim/Global-and-Local-Attention-Based-Free-Form-Image-Inpainting. My fork is located in styler00dollar/Colab-Global-and-Local-Inpainting.
• JiahuiYu/generative_inpainting aka DeepFill v1/v2 with Contextual Attention and Gated Convolution. My fork is located in styler00dollar/Colab-generative-inpainting.
• Video inpainting with Flow-edge Guided Video Completion can be found in vt-vl-lab/FGVC.
  • Alternative with brunomsantiago/FGVC_video_inpaint_colab_drive.
• A lightning Colab that is a combination of a lot of different inpainting networks and loss functions. May also support super resolution in the future, but for now, only inpainting generators are the focus. My Colab is located in styler00dollar/Colab-BasicSR.
• A lightweight genative architexture for image inpainting wih GuardSkill/AptiveGAN. My fork is located in styler00dollar/Colab-AdaptiveGAN.

TTS

• YourTTS: Towards Zero-Shot Multi-Speaker TTS and Zero-Shot Voice Conversion for everyone with Edresson/YourTTS.
  • Demo zero-shot TTS with YourTTS:
  • Demo zero-shot Voice conversion with YourTTS:
• LibriTTS trained multi speaker TTS demo using NVIDIA/flowtron.
• An English female voice (LJSpeech) demo using Rayhane-mamah/Tacotron-2 and r9y9/wavenet_vocoder
• A Mongolian male voice demo using Rayhane-mamah/Tacotron-2 with the Griffin-Lim algorithm.
• An English female voice (LJSpeech) demo using tugstugi/pytorch-dc-tts with the Griffin-Lim algorithm.
• An English female voice (LJSpeech) demo using fatchord/WaveRNN (Tacotron + WaveRNN).
• An English female voice (LJSpeech) demo using mozilla/TTS (Tacotron + WaveRNN).
• NVIDIA/mellotron notebook.
• Voice clone demo using CorentinJ/Real-Time-Voice-Cloning.
• Official ESPnet English/Chinese/Japanese TTS notebook.
• Official ForwardTacotron LJSpeech TTS notebook.
• Text to speech with TalkNet.
  • Inference:
  • Training:

Speech Recognition

• mozilla/DeepSpeech with LM on Youtube videos.
• Wav2Letter+ from NVIDIA/OpenSeq2Seq without LM on Youtube videos.
• Jasper from NVIDIA/OpenSeq2Seq without LM on Youtube videos.
• QuartzNet from NVIDIA/Nemo without LM on Youtube videos.
• QuartzNet from NVIDIA/Nemo without LM with microphone.
• Official ESPnet Spanish->English speech translation notebook.
• English/German/Spain speech recognition with snakers4/silero-models.
• Transcribe your songs with Music-and-Culture-Technology-Lab/omnizart.

Object Detection

• Detecting Twenty-thousand Classes using Image-level Supervision with facebookresearch/Detic.
• Tensorflow object detection: FasterRCNN+InceptionResNet and ssd+mobilenet.
• Cascade RCNN demo using open-mmlab/mmdetection.
• YOLO v3 demo using ayooshkathuria/pytorch-yolo-v3.
• YOLO v4 with AlexeyAB/darknet.
• YOLO v5 with ultralytics/yolov5.
• Object detection on Youtube videos using amdegroot/ssd.pytorch (SSD300).
• CenterNet (Objects as Points) demo using xingyizhou/CenterNet.
• Official DE⫶TR demo notebook facebookresearch/detr.
• Official Google EfficientDet notebook.
• Test and train box-models from Tensorflow detection model zoo with dctian/DeepPiCar.
• Anime object detection with zymk9/yolov5_anime and printcraft/anime-and-cg-characters-detection-using-yolov5. My fork is located in styler00dollar/Colab-yolov5_anime.
• A lightweight vision library for performing large scale object detection/ instance segmentation with obss/sahi.

Segmentation

• Highly Accurate Dichotomous Image Segmentation with xuebinqin/DIS.
• End-to-End Referring Video Object Segmentation with Multimodal Transformers with mttr2021/MTTR.
• SOTA Semantic Segmentation Models in PyTorch with sithu31296/semantic-segmentation.
• Semantic segmentation trained on ADE20K using CSAILVision/semantic-segmentation-pytorch.
• DeepLabV3 from torchvision.
• Fast tracking and segmentation with SiamMask on Youtube videos.
• Real-time semantic segmentation with LightNet++ on Youtube videos.
• Real-time instance segmentation with YOLACT on Youtube videos.
• Instance segmentation with CenterMask.
• Train and test Tensorflow detection model zoo mask models with TannerGilbert/Tensorflow-Object-Detection-API-train-custom-Mask-R-CNN-model. There is also a tutorial dedicated to this repo.
• Open source semantic segmentation toolbox open-mmlab/mmsegmentation.
• Mask RCNN demo using matterport/Mask_RCNN.
• Mask RCNN demo using Detectron.
• Detectron2:
  • Official Detectron2 Mask RCNN demo with facebookresearch/detectron2.
  • A combination of facebookresearch/detectron2, zhanghang1989/detectron2-ResNeSt and youngwanLEE/centermask2 with my fork styler00dollar/Colab-Detectron2:
• Mask RCNN demo from torchvision.
• Example usage of open-mmlab/mmdetection with my fork styler00dollar/Colab-mmdetection.

Pose Detection

• OpenPose on Youtube videos.
• AlphaPose on Youtube videos.
  • v0.2.0:
  • v0.3.0:
• DensePose demo notebook.
• HRNet using lxy5513/hrnet on Youtube videos.
• Keypoint R-CNN from torchvision.

Scene Text Detection

• PixelLink demo notebook.
• Scene text detection using argman/EAST.
• Scene text detection using CRAFT-pytorch.

Image generation

• ruDALL-E fine tuning with "Looking Glass v1.1" by Sber AI and ai_curio. More credits in the notebook.
• Generative Art Using Neural Visual Grammars and Dual Encoders with deepmind/arnheim.
  • 1:
  • 2:
• Generate images from russian texts with sberbank-ai/ru-dalle.
  • With automatic google ENG->RUS translation:
• Generate images from a text prompt with borisdayma/dalle-mini.
• Closed-Form Factorization of Latent Semantics in GANs with genforce/sefa.
• Generate textures with Self-Organising Textures.
• Zero-Shot non-adversarial domain adaptation of pre-trained generators with rinongal/StyleGAN-nada.
• Taming Transformers for High-Resolution Image Synthesis with CompVis/taming-transformers.
  • My fork with smaller Colab is located in styler00dollar/Colab-taming-transformers.
• BigGAN:
  • BigGAN Large Scale GAN Training for High Fidelity Natural Image Synthesis.
  • Text-based image generation with BigGAN and CLIP:
    • Training-Free Text-to-Image Generation with Improved CLIP+GAN Space Optimization with gnobitab/fusedream.
    • Colab by @advadnoun.
      • A modification of that notebook to reduce amount of text/clicks can be found inside styler00dollar/Colab-BigGANxCLIP.
      • Another modification of that Colab by nmkd:
  • Colab by @eyaler / eyaler/clip_biggan.
    • Alternative Colab with CMA-ES is also inside eyaler/clip_biggan.
• StyleGAN2:
  • StyleGAN2 with Differentiable Augmentation with mit-han-lab/data-efficient-gans.
  • Style-based GAN architecture (StyleGAN2) can be found in NVlabs/stylegan2.
  • Generating images from caption and vice versa via CLIP-Guided Generative Latent Space Search with galatolofederico/clip-glass.
  • StyleGAN2 (ADA) with eps696/stylegan2ada.
  • Create StyleGAN2 Steam banners with woctezuma/steam-stylegan2.
  • StyleGAN2-ada with eps696/stylegan2ada.
  • StyleGAN2 with eps696/stylegan2.
• Anime+StyleGAN2:
  • This Anime Does Not Exist - DeepDanbooru Editor by arfa. Original source is here.
  • Style-based GAN architecture (StyleGAN2) with anime face generation.
  • Newest model training attempt by aydao (@AydaoAI) and Colab provided by arfa (@arfafax).
    • That combined with CLIP provided by nagolinc/notebooks. (Warning: Results usually look bad, even had code related errors when I tested it.)
    • A more compact and fixed version of that notebook by me. (Only fixed syntax errors, results still not reliable.)
    • A V2 appeared inside the original repo nagolinc/notebooks, but still has quite a lot of boxes. The psi parameter is also gone.
  • Fine-tuning StyleGAN2 for Cartoon Face Generation with happy-jihye/Cartoon-StyleGan2.
• VQGAN + CLIP by Katherine Crowson.
  • A modification of this notebook found in here and in nerdyrodent/VQGAN-CLIP. Added some explanations, modification, pooling trick and GUI.
  • A notebook based on the crowsonkb notebook. Generates images from text prompts with VQGAN and CLIP (z+quantize method). Source is here and here.
    • A similar notebook, but cleaner and in spanish.
• Further CLIP related notebooks:
  • Styled text-to-drawing synthesis with pschaldenbrand/styleclipdraw.
  • StyleGAN3 with CLIP by nshepperd and Katherine Crowson.
  • Using the VAE from openai/DALL-E and combining it with openai/CLIP. Colab by @advadnoun. (Warning: This is not real DALLE-E, it just uses the official VAE)
  • CLIP + FFT with eps696/aphantasia.
  • Text-guided StyleGAN2 image generation with orpatashnik/StyleCLIP.
  • Text-based image generation with SIREN and CLIP. Original notebook from here.
    • A more compact version can be found inside lucidrains/deep-daze.
    • My version with small modifications to lucidrains/deep-daze. Less printing and Google Drive support with styler00dollar/Colab-deep-daze.
  • Generates images from text prompts with a CLIP conditioned Decision Transformer by Katherine Crowson.
  • Generates images from text prompts with CLIP guided diffusion by crowsonkb.
  • VQGAN to PixelDraw to render the image with dribnet/clipit.
  • Create Disco Diffusion artworks in one line with jina-ai/discoart.
  • Pixel Art Diffusion with KaliYuga-ai/Pixel-Art-Diffusion.
    • Pixel Art Diffusion v3.0 (With Disco Symmetry):
    • Watercolor Diffusion v1.0 by KaliYuga:

Image colorization

• Time-Travel Rephotography with Time-Travel-Rephotography/Time-Travel-Rephotography.github.io.
• DeOldify: A Deep Learning based project for colorizing and restoring old images.
  • My fork that combines deoldify anime and normal deoldify is located in styler00dollar/Colab-DeOldify.
  • deoldify an image (artistic model)
  • deoldify an image (stable model)
  • deoldify an anime image with Dakini/AnimeColorDeOldify.
• Coloring images with pvitoria/ChromaGAN.
• Anime image colorization with reference image. delta6189/Anime-Sketch-Colorizer contains notebooks, but I created a version that does work with Colab which can be found inside styler00dollar/Colab-Anime-Sketch-Colorizer.
• Instance-aware Image Colorization with ericsujw/InstColorization.

Image modification

• DiffusionCLIP: Text-Guided Diffusion Models for Robust Image Manipulation with gwang-kim/DiffusionCLIP.
• Towards Layer-wise Image Vectorization with Picsart-AI-Research/LIVE-Layerwise-Image-Vectorization.
• PyTorch codes for "Towards Robust Blind Face Restoration with Codebook Lookup Transformer" with sczhou/CodeFormer.
• Doing Style Transfer with Progressive Attentional Manifold Alignment with computer-vision2022/PAMA. My fork is located in styler00dollar/Colab-PAMA.
• A Closed-form Solution to Universal Style Transfer with boomb0om/PyTorch-OptimalStyleTransfer.
• Image Style Transfer with a Single Text Condition with paper11667/clipstyler.
• StyleGAN Inversion with HyperNetworks for Real Image Editing with yuval-alaluf/hyperstyle.
  • Inference:
  • Animation:
• Cartoonize faces with bryandlee/animegan2-pytorch.
• StyleGAN of All Trades: Image Manipulation with Only Pretrained StyleGAN with mchong6/SOAT.
• Image-Based CLIP-Guided Essence Transfer with hila-chefer/targetclip.
  • TargetClip:
  • TargetClip + e4e:
• Modidy images with e4e + StyleCLIP with bycloudai/StyleCLIP-e4e-colab.
  • My own version with less boxes and text (Colab-e4e-StyleCLIPglobal):
  • Original Colab:
• Modify images with Mayukhdeb/torch-dreams.
• GAN Prior Embedded Network for Blind Face Restoration in the Wild with yangxy/GPEN. Fork with Colab is in bycloudai/GPEN-colab.
• Towards Real-World Blind Face Restoration with Generative Facial Prior with TencentARC/GFPGAN.
• Discovering Interpretable GAN Controls with harskish/ganspace.
• Image to Sketch with vijishmadhavan/ArtLine.
  • Smooth:
  • Quality:
• Animefy an image with StyleGAN2 and XingruiWang/Animefy.
• Text-guided StyleGAN2 image modification with orpatashnik/StyleCLIP.
• Restore old photos with microsoft/Bringing-Old-Photos-Back-to-Life.
• Decensoring Hentai with Deep Neural Networks. The original repo is deeppomf/DeepCreamPy and my fork is located in styler00dollar/Colab-DeepCreamPy.
• dreamnettech/dreampower is a deep learning algorithm based on DeepNude with the ability to nudify photos of people. My fork is located in styler00dollar/Colab-dreampower.
• Deblurring pictures with TAMU-VITA/DeblurGANv2. My own fork is located in styler00dollar/Colab-DeblurGANv2.
• Deblurring pictures with SeungjunNah/DeepDeblur-PyTorch. My own fork is located in styler00dollar/Colab-DeepDeblur.
• gordicaleksa/pytorch-deepdream will give you the power to create weird and psychedelic-looking images. My fork is located in styler00dollar/Colab-deepdream.
• Single-Image HDR Reconstruction by Learning to Reverse the Camera Pipeline with alex04072000/SingleHDR.
• Watermark removal with vinthony/deep-blind-watermark-removal.
  • A more userfriendly version that does allow custom input with my fork styler00dollar/Colab-deep-watermark.
• Versatile Image-to-Image Translation with linjx-ustc1106/TuiGAN-PyTorch. My Colab is located in styler00dollar/Colab-TuiGAN.
• Image denoising with cszn/DPIR. My colab is located in styler00dollar/Colab-DPIR.
• Blind Face Restoration via Deep Multi-scale Component Dictionaries with csxmli2016/DFDNet.
  • Alternative Colab provided by xinntao/BasicSR.
• Training DFDnet with pytorch and pytorch lightning with styler00dollar/Colab-DFDNet. Uses the fork csxmli2016/DFDNet as a base, which offers more/better code than the official repo csxmli2016/DFDNet.
  • Pytorch:
  • Pytorch lightning (more loss, cleaner code, multi-gpu, etc.):
• Style transfer with linjx-ustc1106/TuiGAN-PyTorch and my fork styler00dollar/Colab-TuiGAN.
• Face editing with mit-han-lab/anycost-gan.
• Remove background with PeterL1n/BackgroundMattingV2.
  • Image:
• Modify attributes of anime faces like eyes, mouth, rotation, etc. with pkhungurn/talking-head-anime-2-demo.
• Reconstruct images and merge multiple images together with chail/latent-composition. My fork is located in styler00dollar/Colab-latent-composition.
• A Residual-Based StyleGAN Encoder via Iterative Refinement with yuval-alaluf/restyle-encoder.
• Navigating the GAN Parameter Space for Semantic Image Editing with yandex-research/navigan.
• A Gated and Bifurcated Stacked U-Net Module for Document Image Dewarping with DVLP-CMATERJU/RectiNet.
• Diverse im2im and vid2vid selfie to anime translation with mchong6/GANsNRoses.
• Age Transformation Using a Style-Based Regression Model with yuval-alaluf/SAM.
  • Inference Notebook:
  • Animation Notebook:

Video modification

• A Google Colab notebook set up for both conventional and machine learning-based video processing. This repo combines VapourSynth and ESRGAN and is located in rlaphoenix/VSGAN.
• Generates a talking face video from an image and an audio using Rudrabha/LipGAN.
• Deoldify a video.
• Decensoring mosaic with HypoX64/DeepMosaics. My own colab fork is located in styler00dollar/Colab-DeepMosaics.
• Inpaint video with vt-vl-lab/FGVC.
• Remove background with PeterL1n/BackgroundMattingV2.
  • Video:

Image classification

• ConvNeXt with facebookresearch/ConvNeXt.
• An end-to-end framework for anime character recognition and tagging with arkel23/animesion. Original Colab by hj is here. My fork is located in styler00dollar/Colab-animesion.
• A combination of a lot of discriminator networks and loss functions for image classification with styler00dollar/Colab-image-classification.
• Image classification with bentrevett/pytorch-image-classification.
  • Multilayer Perceptron:
  • LeNet:
  • AlexNet:
  • VGG:
  • ResNet:
  • My very compact version of ResNet by me with styler00dollar/Colab-image-classification:
• More completely new classification notebooks that are within styler00dollar/Colab-image-classification.
  • ResNeSt:
  • EfficientNet:
  • EfficientNet (lightning):
  • mobilenet_v3:
  • SqueezeNet:
  • RepVGG:
  • vit (lightning):
• Using Autoencoders for unsupervised classification with ardamavi/Unsupervised-Classification-with-Autoencoder. My fork styler00dollar/Colab-UnsupervisedClassification has some improvements and is usable with Colab.
• OpenMMLab Image Classification Toolbox and Benchmark open-mmlab/mmclassification. My fork is located in styler00dollar/Colab-mmclassification.
• Tokens-to-Token ViT: Classification of images with transformers and yitu-opensource/T2T-ViT and my fork styler00dollar/Colab-T2T-ViT.

NLP

• Ai Dungeon alternative with finetuneanon/gpt-neo_dungeon. [Warning: Because of high amount of Google Drive requrests to download the models, there can be errors when you try to use the Colab. It's recommended to either retry another time if you get errors or download the models with the torrent (magnet in Colab) and upload model to your own Drive.]
• Ai Dungeon alternative with KoboldAI/KoboldAI-Client.
• GPT2 English with thecoder-001/GPT-2.
  • A small modification by me to fix usage:
• Finetune GPT2 with ak9250/gpt-2-colab.
• Using russian GPT2 and 3 (based on 2) with sberbank-ai/ru-gpts.
  • Generation with ruGPT3large:
  • Finetuning ruGPT3Small:
• GPT-J-6B Inference Demo with kingoflolz/mesh-transformer-jax.

Misc

Everything that does not really fit into one category

• PyTorch implementation for SDEdit: Image Synthesis and Editing with Stochastic Differential Equations with ermongroup/SDEdit.
• A Video Restoration Transformer with JingyunLiang/VRT.
• Perform image captioning, visual question answering and feature exraction with salesforce/BLIP.
  • My own Colab:
• RuDOLPH: One Hyper-Modal Transformer can be creative as DALL-E and smart as CLIP with sberbank-ai/ru-dolph.
• Text to zooming Video generation with Disco Diffusion v4 by @Somnai_dreams.
• Disco Diffusion v5.4. Credits in the notebook.
• GLIDE: a diffusion-based text-conditional image synthesis model with openai/glide-text2im.
  • text2im:
  • inpaint:
  • clip_guided:
• Pedestrian tracking using ZQPei/deep_sort_pytorch (DeepSORT + YOLOv3).
• Fast and Accurate One-Stage Space-Time Video Super-Resolution with Mukosame/Zooming-Slow-Mo-CVPR-2020. My colab fork is located in styler00dollar/Colab-Zooming-Slow-Mo.
• High-Resolution Image Synthesis with Latent Diffusion Models with CompVis/latent-diffusion. (Text-to-Image, Layout-to-Image, Class-Label-to-Image, Super Resolution, Inpainting, Semantic-Map-to-Image)
• Using CLIP to assosiate pokemon game characteristics with images with CLIPokemon by @ai_curio.
• Simple image captioning model with rmokady/clip_prefix_caption.
• Anime Face Detector using mmdet and mmpose with hysts/anime-face-detector.
• Image Caption Generation Demo with dzryk/antarctic-captions.
• Defocus Blur Detection via Depth Distillation with vinthony/depth-distillation.
• Generate a human 3d model from a 2d picture with facebookresearch/pifuhd.
• Turn a 2D image into a 3D video with ai-coodinator/3D-Photo-Inpainting and vt-vl-lab /3d-photo-inpainting.
• Using Mask-RCNN and ESRGAN to detect bars and mosaic and depixelate content with natethegreate/hent-AI. The official Colab is broken, mine is located in styler00dollar/Colab-hentAI.
• Pretty much the same as hent-AI, but better, with styler00dollar/CenDetect.
• Music Source Separation sigsep/open-unmix-pytorch.
• First Order Motion Model for Image Animation AliaksandrSiarohin/first-order-model.
• Official notebook of 3D Photography using Context-aware Layered Depth Inpainting vt-vl-lab/3d-photo-inpainting.
• Image-GPT notebook.
• Lifespan Age Transformation Synthesis with royorel/Lifespan_Age_Transformation_Synthesis.
• Bayesian Image Reconstruction (inpainting and super resolution) using Deep Generative Models with razvanmarinescu/brgm.
• Contrastive Language-Image Pre-Training with openai/CLIP.
• Real-time View Synthesis with nex-mpi/nex-code.
• FuSta: Hybrid Neural Fusion for Full-frame Video Stabilization with alex04072000/FuSta.
• Defocus Blur Detection via Depth Distillation with vinthony/depth-distillation.
• Boosting Monocular Depth Estimation Models to High-Resolution via Content-Adaptive Multi-Resolution Merging with compphoto/BoostingMonocularDepth.
• Background Matting: The World is Your Green Screen senguptaumd/Background-Matting.
• Taming Visually Guided Sound Generation with v-iashin/SpecVQGAN.
• StyleSDF: High-Resolution 3D-Consistent Image and Geometry Generation with royorel/StyleSDF.
• OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework with OFA-Sys/OFA.
  • Generic Interface:
  • Image Captioning:
  • Referring Expression Comprehension:
  • Open-Domain Visual Question Answering: