Views
No views yet
| Metric | Value |
|---|---|
| Validation Accuracy | 91.67% |
| Macro F1-Score | 0.917 |
| Parameters | 4.05M |
| Model Size | 15.7 MB |
| GPU Latency | 5.81 ms (RTX 4070 SUPER) |
| CPU Latency | 25.76 ms |
| Model | Params (M) | Val Accuracy |
|---|---|---|
| MobileNetV3-Large | 4.24 | 95.45% |
| ResNet50 | 23.58 | 93.94% |
| EfficientNet-B0 | 4.05 | 91.67% |
1import timm
2import torch
3from PIL import Image
4from torchvision import transforms
5
6# Load model
7model = timm.create_model('efficientnet_b0', pretrained=False, num_classes=33)
8state_dict = torch.load('pytorch_model.bin', map_location='cpu')
9model.load_state_dict(state_dict)
10model.eval()
11
12# Preprocessing
13transform = transforms.Compose([
14 transforms.Resize(256),
15 transforms.CenterCrop(224),
16 transforms.ToTensor(),
17 transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
18])
19
20# Inference
21image = Image.open('bacteria_image.jpg').convert('RGB')
22input_tensor = transform(image).unsqueeze(0)
23
24with torch.no_grad():
25 outputs = model(input_tensor)
26 predicted_class = outputs.argmax(dim=1).item()
27
28print(f"Predicted class: {CLASS_NAMES[predicted_class]}")1CLASS_NAMES = [
2 "Acinetobacter_baumannii", "Actinomyces_israelii", "Bacteroides_fragilis",
3 "Bifidobacterium_spp", "Candida_albicans", "Clostridium_perfringens",
4 "Enterococcus_faecalis", "Enterococcus_faecium", "Escherichia_coli",
5 "Fusobacterium", "Lactobacillus_casei", "Lactobacillus_crispatus",
6 "Lactobacillus_delbrueckii", "Lactobacillus_gasseri", "Lactobacillus_jensenii",
7 "Lactobacillus_johnsonii", "Lactobacillus_paracasei", "Lactobacillus_plantarum",
8 "Lactobacillus_reuteri", "Lactobacillus_rhamnosus", "Lactobacillus_salivarius",
9 "Listeria_monocytogenes", "Micrococcus_spp", "Neisseria_gonorrhoeae",
10 "Porphyromonas_gingivalis", "Propionibacterium_acnes", "Proteus",
11 "Pseudomonas_aeruginosa", "Staphylococcus_aureus", "Staphylococcus_epidermidis",
12 "Staphylococcus_saprophyticus", "Streptococcus_agalactiae", "Veillonella"
13]1@inproceedings{hoflaz2025bacterial,
2 title={Lightweight CNNs Outperform Vision Transformers for Bacterial Colony Classification},
3 author={Hoflaz, Ibrahim},
4 booktitle={IEEE Conference},
5 year={2025}
6}