Merge branch 'main' into anonymous

HumanPose/code/loss.py

@@ -1,7 +1,6 @@
 import torch
 from typing import Union
 
-
 from utils.tic import get_positive_definite_matrix, get_tic_covariance
 from models.vit_pose.ViTPose import ViTPose
 from models.stacked_hourglass.StackedHourglass import PoseNet as Hourglass

HumanPose/code/utils/tic.py

@@ -110,6 +110,7 @@ def calculate_ll_per_sample(y_pred: torch.Tensor, precision_hat: torch.Tensor,
 
 def _predictions_hg(hg_level_6: Hourglass, hg_feat: Hourglass,
                     hg_out: Hourglass) -> Callable[[torch.Tensor], torch.Tensor]:
+
     """
     Obtains the model's target predictions for an input.
     This function is used in conjunction with vmap.
@@ -134,7 +135,7 @@ def pred(x: torch.Tensor) -> torch.Tensor:
         return soft_argmax(x).view(x.shape[0], -1)
     return pred
 
-
+  
 def _predictions_vitpose(x: torch.Tensor, imgs: torch.Tensor, vitpose: ViTPose) -> torch.Tensor:
     """
     Obtains the model's target predictions for an input.

README.md

@@ -1,38 +1,36 @@
-![TIC-TAC](https://github.com/meghshukla/TIC-TAC/blob/anonymous/TIC-TAC_gif.gif)
+![TIC-TAC](https://github.com/vita-epfl/TIC-TAC/blob/main/TIC-TAC_gif.gif)
 
 
 # TIC-TAC: A Framework For Improved Covariance Estimation In Deep Heteroscedastic Regression
 
 <a href="https://arxiv.org/abs/2310.18953"><img alt="arXiv" src="https://img.shields.io/badge/arXiv-2310.18953-%23B31B1B?logo=arxiv&logoColor=white"></a>
 <a href="https://www.epfl.ch/labs/vita/heteroscedastic-regression/"><img alt="Project" src="https://img.shields.io/badge/-Project%20Page-lightgrey?logo=Google%20Chrome&color=informational&logoColor=white"></a>
 <a href="https://openreview.net/forum?id=zdNTiTs5gU"><img alt="OpenReview" src="https://img.shields.io/badge/ICML%202024-OpenReview-%236DA252"></a>
-<a href="https://hub.docker.com/repository/docker/meghshukla/tictac/"><img alt="Docker" src="https://img.shields.io/badge/Image-tictac-%232496ED?logo=docker&logoColor=white"></a>
-<br>
-
-
-
 
+<br>
 
 Code repository for "TIC-TAC: A Framework For Improved Covariance Estimation In Deep Heteroscedastic Regression". We address the problem of sub-optimal covariance estimation in deep heteroscedastic regression by proposing a new parameterisation (TIC) and metric (TAC). We derive a new expression, the _Taylor Induced Covariance (TIC)_, which expresses the randomness of the prediction through its gradient and curvature. The _Task Agnostic Correlations (TAC)_ metric leverages the conditioning property of the normal distribution to evaluate the covariance quantitatively.
 
 
 ## Table of contents
 1. [Installation: Docker (recommended) or PIP](#installation)
-1. [Organization](#organization)
-1. [Code Execution](#execution)
-1. [Citation](#citation)
+2. [Organization](#organization)
+3. [Code Execution](#execution)
+4. [Acknowledgement](#acknowledgement)
+5. [Citation](#citation)
 
 
 ## Installation: Docker (recommended) or PIP <a name="installation"></a>
 
-**Docker**: We provide a Docker image which is pre-installed with all required packages. We recommend using this image to ensure reproducibility of our results. Using this image requires setting up Docker on Ubuntu: [Docker](https://docs.docker.com/engine/install/ubuntu/#installation-methods). Once installed, we can use the provided `docker-compose.yaml` file to start our environment with the following command:  `docker-compose run --rm tictac` <br>
+**Docker <a href="https://hub.docker.com/repository/docker/meghshukla/tictac/"><img alt="Docker" src="https://img.shields.io/badge/Image-tictac-%232496ED?logo=docker&logoColor=white"></a>**: We provide a Docker image which is pre-installed with all required packages. We recommend using this image to ensure reproducibility of our results. Using this image requires setting up Docker on Ubuntu: [Docker](https://docs.docker.com/engine/install/ubuntu/#installation-methods). Once installed, we can use the provided `docker-compose.yaml` file to start our environment with the following command:  `docker-compose run --rm tictac` <br>
 
 **PIP**: In case using Docker is not possible, we provide a `requirements.txt` file containing a list of all the packages which can be installed with `pip`.  We recommend setting up a new virtual environment ([link](https://packaging.python.org/en/latest/guides/installing-using-pip-and-virtual-environments/)) and install the packages using:  `pip install -r requirements.txt`
 
 
 ## Organization <a name="organization"></a>
 
 The repository contains four main folders corresponding to the four experiments: `Univariate, Multivariate, UCI`and `HumanPose`. While the first three are self contained, HumanPose requires us to download images corresponding to the [MPII Dataset](https://datasets.d2.mpi-inf.mpg.de/andriluka14cvpr/mpii_human_pose_v1.tar.gz), [LSP Dataset](http://sam.johnson.io/research/lsp.html) and [LSPET Dataset](http://sam.johnson.io/research/lspet.html). For each of these datasets copy-paste all the images `*.jpg` into `HumanPose/data/{mpii OR lsp OR lspet}/images/`. Within `HumanPose`, we have a separate folder `cached`, which holds the generated file `mpii_cache_imgs_{True/False}.npy`. This file stores the post-processed MPII dataset to avoid redundancy every time the code is run.
+Running `python main.py` in the `code` folder executes the code, with configurations specified in `configuration.yml`
 
 
 ## Code Execution <a name="execution"></a>
@@ -44,6 +42,23 @@ Stopping a container once the code execution is complete can be done using:
 1. `docker ps`: List running containers
 2. `docker stop <container id>`
 
-## Acknowledgement 
+## Acknowledgement <a name="acknowledgement"></a>
+
+We thank `https://github.com/jaehyunnn/ViTPose_pytorch` for their implementation of ViTPose which was easily customizable.
+We also borrow [code](https://github.com/meghshukla/ActiveLearningForHumanPose) from the Active Learning for Human Pose library.
+
+## Citation <a name="citation"></a>
+
+If you find this work useful, please consider starring this repository and citing this work!
 
-https://github.com/jaehyunnn/ViTPose_pytorch
+```
+@InProceedings{shukla2024tictac,
+  title = {TIC-TAC: A Framework for Improved Covariance Estimation in Deep Heteroscedastic Regression},
+  author = {Shukla, Megh and Salzmann, Mathieu and Alahi, Alexandre},
+  booktitle = {Proceedings of the 41th International Conference on Machine Learning},
+  year = {2024},
+  series = {Proceedings of Machine Learning Research},
+  month = {21--27 Jul},
+  publisher = {PMLR}
+}
+```

UCI/UCI.py

@@ -30,7 +30,7 @@
 
 # Configuration
 trials = 10
-dataset_uci = 'concrete'
+dataset_uci = 'naval'
 # Possible options: red_wine, white_wine, energy, concrete, power, air, naval, electrical
 # abalone, gas_turbine, appliances, parkinson
 

Univariate/univariate.py

@@ -36,12 +36,12 @@
 
 ###### Configuration ######
 # Sinusoidal Configuration
-varying_amplitude = False
-invert_varying = None
+varying_amplitude = True
+invert_varying = False
 frequency = 2 * np.pi * 1
 
 # Create experiment folder
-experiment_name = 'Results/VaryingAmplitude_{}_and_InvertVarying_{}'.format(
+experiment_name = 'ICML_Results/VaryingAmplitude_{}_and_InvertVarying_{}'.format(
     varying_amplitude, invert_varying)