GUguifortaine·Apr 10, 2016Submission
Deeper LSTM+ Normalized CNN for Visual Question Answering
4 points · Open discussionGT-Vision-Lab
VQA_LSTM_CNN
Lua2016–2016
GT-Vision-Lab / VQA_LSTM_CNN
VQA_LSTM_CNN
Train a deeper LSTM and normalized CNN Visual Question Answering model. This current code can get 58.16 on OpenEnded and 63.09 on Multiple-Choice on test-standard.
2Distinct people
2Mentions
0Explicit recommenders
386GitHub stars
What to know before exploring
What it is
Train a deeper LSTM and normalized CNN Visual Question Answering model. This current code can get 58.16 on OpenEnded and 63.09 on Multiple-Choice on test-standard.
GitHub
Where it is used
No documented product use has been verified yet.
What people are saying
No recommendation phrase was found in the indexed comments.
No classified warning is attached to this page. Read the discussions before treating silence as approval.
Also in these discussions
No other repository is repeatedly linked in the same discussions.
This week on Hacker News
Nov 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 5:04:57 PM UTC.
No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.
Mentions over time
HN mentions by 6-hour interval2 dated mentions
Observed mentions: 2016-04-06 00:29:34 UTC to 2016-04-10 14:20:17 UTC.
Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.
View mention counts and date ranges
| Period | Start (UTC) | End (exclusive, UTC) | Mentions | Submissions | Comments |
|---|---|---|---|---|---|
| Apr 6, 00:00 | 2016-04-06 00:00:00 UTC | 2016-04-06 06:00:00 UTC | 1 | 0 | 1 |
| Apr 6, 06:00 | 2016-04-06 06:00:00 UTC | 2016-04-06 12:00:00 UTC | 0 | 0 | 0 |
| Apr 6, 12:00 | 2016-04-06 12:00:00 UTC | 2016-04-06 18:00:00 UTC | 0 | 0 | 0 |
| Apr 6, 18:00 | 2016-04-06 18:00:00 UTC | 2016-04-07 00:00:00 UTC | 0 | 0 | 0 |
| Apr 7, 00:00 | 2016-04-07 00:00:00 UTC | 2016-04-07 06:00:00 UTC | 0 | 0 | 0 |
| Apr 7, 06:00 | 2016-04-07 06:00:00 UTC | 2016-04-07 12:00:00 UTC | 0 | 0 | 0 |
| Apr 7, 12:00 | 2016-04-07 12:00:00 UTC | 2016-04-07 18:00:00 UTC | 0 | 0 | 0 |
| Apr 7, 18:00 | 2016-04-07 18:00:00 UTC | 2016-04-08 00:00:00 UTC | 0 | 0 | 0 |
| Apr 8, 00:00 | 2016-04-08 00:00:00 UTC | 2016-04-08 06:00:00 UTC | 0 | 0 | 0 |
| Apr 8, 06:00 | 2016-04-08 06:00:00 UTC | 2016-04-08 12:00:00 UTC | 0 | 0 | 0 |
| Apr 8, 12:00 | 2016-04-08 12:00:00 UTC | 2016-04-08 18:00:00 UTC | 0 | 0 | 0 |
| Apr 8, 18:00 | 2016-04-08 18:00:00 UTC | 2016-04-09 00:00:00 UTC | 0 | 0 | 0 |
| Apr 9, 00:00 | 2016-04-09 00:00:00 UTC | 2016-04-09 06:00:00 UTC | 0 | 0 | 0 |
| Apr 9, 06:00 | 2016-04-09 06:00:00 UTC | 2016-04-09 12:00:00 UTC | 0 | 0 | 0 |
| Apr 9, 12:00 | 2016-04-09 12:00:00 UTC | 2016-04-09 18:00:00 UTC | 0 | 0 | 0 |
| Apr 9, 18:00 | 2016-04-09 18:00:00 UTC | 2016-04-10 00:00:00 UTC | 0 | 0 | 0 |
| Apr 10, 00:00 | 2016-04-10 00:00:00 UTC | 2016-04-10 06:00:00 UTC | 0 | 0 | 0 |
| Apr 10, 06:00 | 2016-04-10 06:00:00 UTC | 2016-04-10 12:00:00 UTC | 0 | 0 | 0 |
| Apr 10, 12:00 | 2016-04-10 12:00:00 UTC | 2016-04-10 18:00:00 UTC | 1 | 1 | 0 |
What Hacker News says
2 sourcesThanks for your work on this, I find the VQA task really interesting.
The classification-based approach is definitely the part I find unsatisfying about this task. The problem to me is that it biases the models learned very strongly towards the data that was collected for training and testing.
Has anyone tried outputting a vector from the model, and using cosine to predict the nearest word/phrase/sentence etc? This seems to work for non-visual QA.[1] Training is performed using noise contrastive estimation. I've discussed this idea with the Virginia Tech team, but I haven't had time to try it, and they seemed a little skeptical.[2]
[1] https://cs.umd.edu/~miyyer/qblearn/
[2] https://github.com/VT-vision-lab/VQA_LSTM_CNN/issues/14
In Deep learning for visual question answering: demo with Keras code