Files
gobot/board-vision/src/model.py
2025-01-05 18:36:42 +01:00

50 lines
1.3 KiB (Stored with Git LFS)
Python

import torch
import torch.nn.functional as F
import torch.nn as nn
import numpy as np
TARGET_SIZE = (25, 25)
class Net(nn.Module):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.conv1 = nn.Conv2d(1, 12, 5) # 24x24 -> 21x21
self.maxpool = nn.MaxPool2d(2, 2) # 21x21 -> 10x10
self.conv2 = nn.Conv2d(12, 16, 3) # 10x10 -> 8x8
self.conv3 = nn.Conv2d(16, 32, 3) # 8x8 -> 6x6
self.fc1 = nn.Linear(32 * 6 * 6, 120)
self.fc2 = nn.Linear(120, 48)
self.fc3 = nn.Linear(48, 2)
def forward(self, x):
x = F.relu(self.conv1(x))
x = self.maxpool(x)
x = F.relu(self.conv2(x))
x = F.relu(self.conv3(x))
#print(x.shape)
x = torch.flatten(x, 1)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = F.softmax(self.fc3(x))
return x
def classify_stone_empty(s: list[np.ndarray], model: nn.Module) -> bool:
s = torch.Tensor(s)
s = s[:,:,:,0]
s = s[:,torch.newaxis,:,:].float() / 255.0
s = s.to(0)
return torch.argmax(model(s), dim=1)
def load_model(f: str) -> torch.nn.Module:
torch.autograd.set_grad_enabled(False)
model = Net()
model.load_state_dict(torch.load(f, weights_only=True))
model.eval()
model.to(0)
return model