google-deepmind/kinetics-i3d issues and pull requests

#102 - flow_imagenet checkpoint

Issue - State: open - Opened by nik123 over 4 years ago

#101 - Pretrained weights on ImageNet only?

Issue - State: closed - Opened by kentwhf over 4 years ago

#100 - extract features

Issue - State: open - Opened by Serendipity-LC over 4 years ago - 5 comments

#99 - Using I3D as a feature extractor for anomaly detection task

Issue - State: closed - Opened by hongsamvo over 4 years ago - 2 comments

#98 - How to train my i3d model

Issue - State: open - Opened by lixue17868800524 over 4 years ago - 3 comments

#97 - Train on videos of different length

Issue - State: closed - Opened by kentwhf over 4 years ago - 1 comment

#96 - I3D on other dataset

Issue - State: open - Opened by kayleeliyx over 4 years ago - 4 comments

#95 - Is there a checkpoint pre-trained only on ImageNet

Issue - State: open - Opened by danielkorat over 4 years ago

#94 - Running on Composite Video file

Issue - State: closed - Opened by harshsp31 over 4 years ago - 2 comments

#93 - Ideas to improve the training accuracy

Issue - State: open - Opened by Leslie-Fang over 4 years ago - 4 comments

#92 - Allocation of 513802240 exceeds 10% of system memory.

Issue - State: open - Opened by Nadern96 over 4 years ago - 1 comment

#91 - AttributeError: module 'tensorflow._api.v1.distribute' has no attribute 'experimental'

Issue - State: open - Opened by tempdeltavalue over 4 years ago - 1 comment

#90 - Frame Numbers in Trained Model?

Issue - State: open - Opened by artest08 almost 5 years ago

#89 - Difference between provided checkpoints.

Issue - State: open - Opened by QwCao almost 5 years ago - 2 comments

#88 - Successor network

Issue - State: closed - Opened by ir0nt0ad almost 5 years ago - 2 comments

#87 - Video preprocessing code

Issue - State: open - Opened by javithe7 almost 5 years ago - 5 comments

#86 - Trouble saving I3d features

Issue - State: open - Opened by GayatriPurandharT almost 5 years ago - 2 comments

#85 - Tensorflow and Sonnet versions 2.0 doesn't work

Issue - State: open - Opened by NagabhushanSN95 almost 5 years ago - 1 comment

#84 - Can you update the "requirements.txt"?

Issue - State: closed - Opened by VijayRameshkumar almost 5 years ago - 2 comments

#83 - Could I share your well-trained models on UCF-101 and HMDB-51 (RGB and Optical flow)

Issue - State: open - Opened by junmuzi almost 5 years ago

#82 - training from scratch cost?

Issue - State: open - Opened by LindseyScharpman almost 5 years ago

#81 - Dependency with cuda 9.0

Issue - State: closed - Opened by zeynepgokce almost 5 years ago - 1 comment

#80 - how to deal with long video and short video

Issue - State: open - Opened by zouying-sjtu almost 5 years ago - 1 comment

#79 - File format and pre-processing of RGB/Flow data

Issue - State: closed - Opened by daveboat almost 5 years ago - 9 comments

#78 - Has anyone tried to run this on mobile?

Issue - State: open - Opened by stev3 almost 5 years ago - 1 comment

#77 - How should I normalize the framerate?

Issue - State: closed - Opened by skaws2003 about 5 years ago - 2 comments

#76 - Can you tell me the hyperparameters in hmdb51?

Issue - State: open - Opened by seominseok0429 about 5 years ago - 2 comments

#75 - I3D crashes at the last average pooling 3D if the network has a shape < 2 in its temporal dimension

Issue - State: closed - Opened by mourad1081 about 5 years ago - 2 comments

#74 - ResourceExhaustedError when running the file evaluate_sample.py with my own video

Issue - State: closed - Opened by mourad1081 about 5 years ago - 3 comments

#73 - Can you provide the score files as well as the training and testing lists on hmdb51 and ucf101 ?

Issue - State: open - Opened by ZealACMer about 5 years ago

#72 - about frames

Issue - State: open - Opened by ghost about 5 years ago - 1 comment

#71 - How to select clip temporally?

Issue - State: closed - Opened by hollowgalaxy about 5 years ago - 1 comment

#70 - out of memory

Issue - State: open - Opened by geminikq about 5 years ago - 1 comment

#69 - what are input_arrays and output_arrays for the model

Issue - State: open - Opened by samaa about 5 years ago

#68 - Difference in precision of the probabilities & logits obtained

Issue - State: open - Opened by sparsh-b about 5 years ago

#67 - Problem with dependencies

Issue - State: closed - Opened by v-kogr about 5 years ago - 7 comments

#66 - extract features

Issue - State: open - Opened by 123liluky about 5 years ago - 1 comment

#65 - What minimum amount of GPU is needed to finetune pretrained ? 8gb GPU is enough? or i need 12 gb GPU?

Issue - State: open - Opened by siddharth2022 over 5 years ago - 3 comments

#64 - What is the difference of I3d and 3d ConvNet?

Issue - State: open - Opened by cemilzalluhoglu over 5 years ago - 9 comments

#63 - Localization of Activity

Issue - State: open - Opened by pcarun27 over 5 years ago - 2 comments

#62 - How to use the kinectics pre-trained I3D as Feature Extractor for my own data

Issue - State: closed - Opened by rishabh2301 over 5 years ago - 1 comment

#61 - how to test model on my sample video

Issue - State: open - Opened by alighofrani95 over 5 years ago - 4 comments

#60 - Fusion of RGB and flow models?

Issue - State: closed - Opened by ir0nt0ad over 5 years ago - 1 comment

#59 - pre-process videos to 25 FPS

Issue - State: open - Opened by bomri over 5 years ago - 6 comments

#58 - Correct links to Kinetics dataset

Pull Request - State: open - Opened by anuragarnab over 5 years ago - 3 comments

#57 - how to use the pretrained model

Issue - State: open - Opened by ilovekj over 5 years ago

#56 - Kinetics-600 checkpoints for optical flow

Issue - State: closed - Opened by gaurvigoyal over 5 years ago - 4 comments

#55 - META DATA FILES OF UCF-101 DATASET FOR ACTION RECOGNITION

Issue - State: open - Opened by abhisheksurya578 over 5 years ago

#54 - where can I get the .pb file for the model

Issue - State: open - Opened by ajcwhiteintel over 5 years ago

#53 - The difference between the 3D Inception Module and the 2D.

Issue - State: open - Opened by zeal-up over 5 years ago

#52 - code to load 2D parameters (pretrained on imagenet) to 3D model

Issue - State: open - Opened by emergencyd over 5 years ago - 3 comments

#51 - Missing data in Kinetics 400

Issue - State: open - Opened by jhagege over 5 years ago - 1 comment

#48 - Dependencies

Issue - State: open - Opened by zwenaing over 5 years ago - 1 comment

#47 - 5k steps on UCF-101 => steps in terms of?

Issue - State: open - Opened by MStumpp over 5 years ago - 4 comments

#46 - train from scratch on ucf101 dataset

Issue - State: open - Opened by leviswind over 5 years ago - 10 comments

#45 - train on my own dataset

Issue - State: open - Opened by 95xueqian over 5 years ago - 14 comments

#44 - How to test using the whole videos?

Issue - State: closed - Opened by zeal-up over 5 years ago - 4 comments

#43 - What is the minimum hardware requirement to fine-tune I3D on new dataset?

Issue - State: closed - Opened by stephencheung16 over 5 years ago - 3 comments

#42 - depth data

Issue - State: open - Opened by sanaz-rezaei almost 6 years ago - 1 comment

#41 - LiteFlowNet over OpenCV TV-L1 optical flow algorithm. Can we ?

Issue - State: open - Opened by estathop almost 6 years ago - 5 comments

#40 - Feature Extraction from Last Global Average Pooling Layer

Issue - State: closed - Opened by estathop almost 6 years ago - 7 comments

#39 - Regarding the 2 dimensions of the Optical Flow

Issue - State: open - Opened by cdjkim almost 6 years ago - 2 comments

#38 - Has anybody reproduced the results on UCF101 and HMDB

Issue - State: open - Opened by XudongLinthu almost 6 years ago - 2 comments

#37 - Instructions how to train the model?

Issue - State: open - Opened by jillelajitta almost 6 years ago

#36 - Regarding TV-L1 Optical Flow

Issue - State: closed - Opened by november07 almost 6 years ago - 4 comments

#35 - is there have pytorch version?

Issue - State: open - Opened by cuixing158 almost 6 years ago - 1 comment

#34 - Data-preprocessing for kinetics-400

Issue - State: open - Opened by YongyiTang92 almost 6 years ago - 17 comments

#33 - Parameters for finetuning on UCF101 and JHMDB

Issue - State: open - Opened by jiaozizhao almost 6 years ago - 1 comment

#32 - How can l adapt evaluate_sample.py for hundred of samples ?

Issue - State: open - Opened by Benjiou about 6 years ago - 1 comment

#31 - no

Issue - State: open - Opened by pinkfloyd06 about 6 years ago

#30 - *.npy files include RGB and flow ，which include 79 frames respecitvely.but each rgb frame to a flow frame?

Issue - State: open - Opened by daixiangzi about 6 years ago

#29 - automatic sign language recognition.

Issue - State: open - Opened by AtaaEddin about 6 years ago - 2 comments

#28 - Is it better to train from scratch on Kinetics-600?

Issue - State: open - Opened by WuJunhui about 6 years ago - 36 comments

#27 - What regularizers do you use when training?

Issue - State: open - Opened by WuJunhui about 6 years ago - 1 comment

#26 - How to generate .npy files for different videos?

Issue - State: open - Opened by AdarshMJ about 6 years ago - 3 comments

#25 - RGB or BGR ?

Issue - State: open - Opened by liuqinying about 6 years ago - 2 comments

#24 - could you release the code to extract optical flow?

Issue - State: open - Opened by hjeun about 6 years ago - 2 comments

#23 - about batch size

Issue - State: open - Opened by WuJunhui over 6 years ago - 4 comments

#22 - How can I train my own model?

Issue - State: open - Opened by DeepStillWater over 6 years ago - 6 comments

#21 - Shape of *.npy file?

Issue - State: open - Opened by egg347 over 6 years ago - 1 comment

#20 - What is the procedure for fine-tuning the model on another dataset?

Issue - State: open - Opened by PoorvaRane over 6 years ago

#19 - IDT Handcrafted Features

Issue - State: open - Opened by moaljazaery over 6 years ago

#18 - What is the nature of deep convnet used in i3d, is it a vgg16 ?

Issue - State: open - Opened by pinkfloyd06 over 6 years ago - 2 comments

#17 - What is the code used for optical flow ?

Issue - State: open - Opened by pinkfloyd06 over 6 years ago

#16 - Can we use i3d on ucf101

Issue - State: open - Opened by bupt-zsp over 6 years ago - 2 comments

#15 - Update README.md

Pull Request - State: open - Opened by petergehler almost 7 years ago - 1 comment

#14 - Where can l dowload the pre-trained model on kinetics dataset?

Issue - State: open - Opened by pinkfloyd06 almost 7 years ago - 2 comments

#13 - Image preprocessing

Issue - State: open - Opened by seann999 almost 7 years ago - 9 comments

#12 - Where to get the pre-trained model on kinetics dataset?

Issue - State: closed - Opened by wanjinchang almost 7 years ago - 6 comments

#11 - I think it is typo

Pull Request - State: closed - Opened by rimchang almost 7 years ago - 1 comment

#10 - How can I apply the first N layers of the model to a video file?

Issue - State: closed - Opened by rcrespocano almost 7 years ago - 1 comment

#9 - Generating Optical Flow

Issue - State: open - Opened by prakashjayy almost 7 years ago - 2 comments

#8 - Value for dropout?

Issue - State: open - Opened by cesarsouza almost 7 years ago

#7 - Could you provide the I3D model of CAFFE version

Issue - State: open - Opened by baiyancheng20 almost 7 years ago

#6 - fine-tuned model on UCF101 and HMDB51

Issue - State: open - Opened by shuangshuangguo almost 7 years ago

#5 - extract TV-L1 from HMDB-51

Pull Request - State: open - Opened by mlakhal about 7 years ago - 1 comment

#4 - Minor fixes

Pull Request - State: open - Opened by chuckcho about 7 years ago

#3 - 3D Inception v1 Model Train from Scratch on UCF101 Data Set

Issue - State: open - Opened by ahkarami about 7 years ago - 9 comments

#2 - Will the training code be released in the future?

Issue - State: open - Opened by gooners1886 about 7 years ago - 11 comments

#1 - Finetune the pretrained model on UCF101

Issue - State: closed - Opened by vra about 7 years ago - 15 comments

GitHub / google-deepmind/kinetics-i3d issues and pull requests