README.md 4.81 KB
Newer Older
1
2
3
4
5
6
7
8
# TensorFlow Research Models

This folder contains machine learning models implemented by researchers in
[TensorFlow](https://tensorflow.org). The models are maintained by their
respective authors. To propose a model for inclusion, please submit a pull
request.

## Models
Andre Araujo's avatar
Andre Araujo committed
9
10
11
12
13
14
15
16

-   [adversarial_crypto](adversarial_crypto): protecting communications with
    adversarial neural cryptography.
-   [adversarial_text](adversarial_text): semi-supervised sequence learning with
    adversarial training.
-   [attention_ocr](attention_ocr): a model for real-world image text
    extraction.
-   [audioset](audioset): Models and supporting code for use with
krantideep95's avatar
krantideep95 committed
17
    [AudioSet](http://g.co/audioset).
Andre Araujo's avatar
Andre Araujo committed
18
-   [autoencoder](autoencoder): various autoencoders.
19
-   [brain_coder](brain_coder): Program synthesis with reinforcement learning.
Andre Araujo's avatar
Andre Araujo committed
20
21
22
23
24
-   [cognitive_mapping_and_planning](cognitive_mapping_and_planning):
    implementation of a spatial memory based mapping and planning architecture
    for visual navigation.
-   [compression](compression): compressing and decompressing images using a
    pre-trained Residual GRU network.
Kevin Clark's avatar
Kevin Clark committed
25
26
-   [cvt_text](cvt_text): semi-supervised sequence learning with cross-view
    training.
Carlos Riquelme's avatar
Carlos Riquelme committed
27
-   [deep_contextual_bandits](deep_contextual_bandits): code for a variety of contextual bandits algorithms using deep neural networks and Thompson sampling.
28
-   [deep_speech](deep_speech): automatic speech recognition.
Anya Petrova's avatar
Anya Petrova committed
29
-   [deeplab](deeplab): deep labeling for semantic image segmentation.
Andre Araujo's avatar
Andre Araujo committed
30
-   [delf](delf): deep local features for image matching and retrieval.
31
32
-   [differential_privacy](differential_privacy): differential privacy for training
    data.
Andre Araujo's avatar
Andre Araujo committed
33
-   [domain_adaptation](domain_adaptation): domain separation networks.
Dieterich Lawson's avatar
Dieterich Lawson committed
34
35
-   [fivo](fivo): filtering variational objectives for training generative
    sequence models.
36
-   [gan](gan): generative adversarial networks.
Andre Araujo's avatar
Andre Araujo committed
37
38
-   [im2txt](im2txt): image-to-text neural network for image captioning.
-   [inception](inception): deep convolutional networks for computer vision.
39
-   [keypointnet](keypointnet): discovery of latent 3D keypoints via end-to-end
Mohammad Norouzi's avatar
.  
Mohammad Norouzi committed
40
    geometric eeasoning [[demo](https://keypointnet.github.io/)].
Andre Araujo's avatar
Andre Araujo committed
41
42
-   [learning_to_remember_rare_events](learning_to_remember_rare_events): a
    large-scale life-long memory module for use in deep learning.
43
44
-   [learning_unsupervised_learning](learning_unsupervised_learning): a
    meta-learned unsupervised learning update rule.
Chris Waterson's avatar
Chris Waterson committed
45
46
-   [lexnet_nc](lexnet_nc): a distributed model for noun compound relationship
    classification.
Andre Araujo's avatar
Andre Araujo committed
47
48
49
-   [lfads](lfads): sequential variational autoencoder for analyzing
    neuroscience data.
-   [lm_1b](lm_1b): language modeling on the one billion word benchmark.
Hoang Trieu Trinh's avatar
Hoang Trieu Trinh committed
50
-   [lm_commonsense](lm_commonsense): commonsense reasoning using language models.
Andrew M. Dai's avatar
Andrew M. Dai committed
51
-   [maskgan](maskgan): text generation with GANs.
Andre Araujo's avatar
Andre Araujo committed
52
53
54
55
56
57
58
59
60
61
62
-   [namignizer](namignizer): recognize and generate names.
-   [neural_gpu](neural_gpu): highly parallel neural computer.
-   [neural_programmer](neural_programmer): neural network augmented with logic
    and mathematic operations.
-   [next_frame_prediction](next_frame_prediction): probabilistic future frame
    synthesis via cross convolutional networks.
-   [object_detection](object_detection): localizing and identifying multiple
    objects in a single image.
-   [pcl_rl](pcl_rl): code for several reinforcement learning algorithms,
    including Path Consistency Learning.
-   [ptn](ptn): perspective transformer nets for 3D object reconstruction.
63
-   [marco](marco): automating the evaluation of crystallization experiments.
Andre Araujo's avatar
Andre Araujo committed
64
65
66
67
68
69
-   [qa_kg](qa_kg): module networks for question answering on knowledge graphs.
-   [real_nvp](real_nvp): density estimation using real-valued non-volume
    preserving (real NVP) transformations.
-   [rebar](rebar): low-variance, unbiased gradient estimates for discrete
    latent variable models.
-   [resnet](resnet): deep and wide residual networks.
Akosua Busia's avatar
Akosua Busia committed
70
71
-   [seq2species](seq2species): deep learning solution for read-level taxonomic
    classification.
Andre Araujo's avatar
Andre Araujo committed
72
73
74
75
76
77
78
-   [skip_thoughts](skip_thoughts): recurrent neural network sentence-to-vector
    encoder.
-   [slim](slim): image classification models in TF-Slim.
-   [street](street): identify the name of a street (in France) from an image
    using a Deep RNN.
-   [swivel](swivel): the Swivel algorithm for generating word embeddings.
-   [syntaxnet](syntaxnet): neural models of natural language syntax.
Corey Lynch's avatar
Corey Lynch committed
79
-   [tcn](tcn): Self-supervised representation learning from multi-view video.
Andre Araujo's avatar
Andre Araujo committed
80
81
82
83
-   [textsum](textsum): sequence-to-sequence with attention model for text
    summarization.
-   [transformer](transformer): spatial transformer network, which allows the
    spatial manipulation of data within the network.
Reza Mahjourian's avatar
Reza Mahjourian committed
84
85
-   [vid2depth](vid2depth): learning depth and ego-motion unsupervised from
    raw monocular video.
Andre Araujo's avatar
Andre Araujo committed
86
87
-   [video_prediction](video_prediction): predicting future video frames with
    neural advection.