# CSE144 Homework4-part2 Image Classification — Data

来源：https://www.kaggle.com/competitions/cse144-hw4-part2/data

采集时间：2026-10-19T11:07:03Z

The data is a subsample of[ ImageNet](https://ieeexplore.ieee.org/document/5206848). 

### Files
There are four files in the dataset: `train.pkl` and `test.pkl`, `helper.ipynb`, and `sample_submission.csv`.

* `helper.ipynb`: code snippets for you to read the train and test data with pickle, and to display an image. You need to use Python3. 

* `train.pkl`: training data serialized with pickle. there are 10000 samples and 10000 corresponding labels. 

* `test.pkl`: test data serialized with pickle. there are 2000 samples, no label. 

* `sample_submission.csv`: to submit your predictions, you should use the exact same format as this. (except for changing the numbers in the label column to the one that's output from your model). The file should contain two columns: id and Label. id is the index of the test data, starting from 0 to 1999 for 2000 test samples. Label is corresponding classification label of that data, which is an integer from 1 to 10.
The file should contain a header and have the following format:
> id,label
0,1
1,3
2,10
3,2
...
1999,5


### Data Format
`train.pkl`, after unpickling, is a dictionary with "data" and "label" fields. `test.pkl` contains just "data", no "label". There are 10000 samples in train and 2000 samples in test. 

"data"
> numpy array with samples x 12288. For each sample, first 4096 numbers represent red channel, next 4096 numbers green channel, last 4096 numbers represent blue channel.

"label"
> number representing image class, indexing starts at 1 and ends at 10. Here are the ten classes:
 1: 'dog',
 2: 'rabbit',
 3: 'deer',
 4: 'elephant',
 5: 'horse',
 6: 'squirrel',
 7: 'tiger',
 8: 'wolf',
 9: 'chimpanzee',
 10: 'panda'.