# Overview
This challenge concerns the automated recognition of visual objects in images using machine learning techniques. The dataset you will work including 23 types of objects in realistic scenes, where training samples and their ground truth are provided. You will develop a machine learning model that will make predictions on unseen test data. The performance will be measured upon submission of the prediction results. A leaderboard will be published after each submission. **You need to use your URN as the Team Name** in the Kaggle page. **Otherwise, your score won't be recorded.** Multiple submissions are encouraged if you wish to improve your models. The latest model for each student will be used for the assessment. You may save multiple prediction results and choose the best one for the final submission.

You can use Swin-Transformer in this week's lab sheet to do this challenge.

# Important Rule
You need to use your URN as the Team Name in the Kaggle page. Otherwise, your score won't be recorded.

# Code Submission
You need to submit your code to Surreylearn before the code submission due date (see SurreyLearn for further detail)

# Participation
Please click the join competition button on the page to join it. Otherwise, you cannot submit your result.


# Task 
### Overview
This is a dataset about **Visual Object Classification**. The goal of this challenge is to recognize objects from a number of visual object classes in realistic scenes. There are twenty-three object classes in the datasets:
* person, hand, foot, head
* bird, cat, cow, dog, horse, sheep
* aeroplane, bicycle, boat, bus, car, motorbike, train
* bottle, chair, dining table, potted plant, sofa, tv/monitor


### Dataset Size
- Training set: 5011 images
- Testing set: 4952 images
### Training and Testing Data
We will provide the images and labels for the training set and images for the testing set.

# Evaluation
We will evaluate your prediction result on the testing set. So you will need to submit the result on this competition page. **The minimum requirement of this challenge is 1.0 Mean ColumnWise Log Loss (lower result indicate better performance).**
 
# Submission Format
Your submission should be a csv file, with 24 columns in it, the first is the id of the test image and the second to 24th are the predicted probability of each visual object class (e.g., person 0.49, cat 0.97, dog 0.01). We will provide an example of the submission file in the Data Section.