Skip to content

Commit

Permalink
Updated readme image
Browse files Browse the repository at this point in the history
  • Loading branch information
rickycorte committed Jun 13, 2023
1 parent d75edda commit 2448c06
Showing 1 changed file with 7 additions and 7 deletions.
14 changes: 7 additions & 7 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,8 @@

Simple segmentation model using [Unet architecture](https://arxiv.org/abs/1505.04597) able to predict binary masks of Pokemons trading cards. Unfortunately, at the moment no pokemon segmentation dataset is available. The model is trained on a custom dataset generated from cards' foil effects.

**At the moment the dataset used in this repository is not publicly available.**

![prediction example](misc/predict_example.jpg)
<p align = "center">
Figure 1. Example prediction of cards that the model never saw during training.
Expand All @@ -12,8 +14,8 @@ Figure 1. Example prediction of cards that the model never saw during training.
Clone and install required dependencies. Its recommended to use conda to create a python environment.

```
git clone <>
cd <>
git clone https://github.com/rickycorte/pokemon-card-segmentation
cd pokemon-card-segmentation
conda create -n pkm python=3.11
conda activate pkm
Expand All @@ -36,14 +38,14 @@ to generate masks and overlapped images inside `output` folder.

Since no dataset is available and creating a decent amount of labels manually would require a lot of time, I developed a simple script that generates reasonable labels starting from card foils scrapped from the internet.

![foils examples](misc/predict_example.jpg)
![foils examples](misc/foils.png)
<p align = "center">
Figure 2. Types of card foil effects.
</p>

As you can see from the image above, a good amount of foil effects are pretty hard to convert into actual segmentation labels especially on newer cards (see Figure 1) where the Pokemon uses almost all the space available. In some cases, this type of card can generate acceptable labels.
As you can see from the image above, a good amount of foil effects are pretty hard (Figure 2, first two images)to convert into actual segmentation labels especially on newer cards (see Figure 1) where the Pokemon uses almost all the space available. In some cases, this type of card can generate acceptable labels.

At the moment, I used only regular cards whereas the Pokemon uses only half of the space to train the model. This means that the predictions generated on cards like Figure only rely on the generalization capabilities of the model.
At the moment, I used only regular cards whereas the Pokemon uses only half (Figure 2, third image)of the space to train the model. This means that the predictions generated on cards like Figure only rely on the generalization capabilities of the model.

Labels generated from newer cards, despite being somewhat acceptable, tend to be really noisy and confuse the model leading to poor generalization on unseen cards.

Expand All @@ -53,5 +55,3 @@ Figure 3. Example of the generated label for a card starting from its foil effec
</p>

The data pre-processing can be found in `data_preprocess.py`.


0 comments on commit 2448c06

Please sign in to comment.