vector length for each level is not the same

[leizhaocs]

Hi, thank you for your implementation.

Based on my understanding of the paper, the length of the vector to each level in figure 1 of the paper should be the same. According to the paper "The estimated class distribution forms a class vector, which is then concatenated with the original feature vector to be input to the next level of cascade".

However, in your implementation, the input to the first level is 1520, the input to the second level is 1560, the input to the third level is 1600, etc.

If I misunderstood the paper, please kindly tell me.

Anyway, I fixed the code myself and find no significant difference in accuracy.

[leopiney]

Hi @shuzihua

You're definitely right. The paper is very clear explaining that part of the algorithm now that I see.
Even though there's no noteworthy difference in the results, feel free to collaborate on this repository!

Thanks!