|
|
|
import pandas as pd |
|
import numpy as np |
|
import os |
|
from tqdm import tqdm |
|
import timm |
|
import torchvision.transforms as T |
|
import timm,wandb,albumentations as A |
|
from albumentations.pytorch import ToTensorV2 |
|
|
|
from PIL import Image |
|
import torch |
|
import torch.nn as nn |
|
import json |
|
|
|
|
|
|
|
|
|
|
|
|
|
SZ = 224 |
|
LABELS = json.load(open("./labels_class_map_rev.json")) |
|
ORIGINAL_LABELS = json.load(open("./original_mapping.json")) |
|
def is_gpu_available(): |
|
"""Check if the python package `onnxruntime-gpu` is installed.""" |
|
return torch.cuda.is_available() |
|
|
|
VALID_AUG = A.Compose([ |
|
A.SmallestMaxSize(max_size=SZ + 16, p=1.0), |
|
A.CenterCrop(height=SZ, width=SZ, p=1.0), |
|
A.Normalize(), |
|
ToTensorV2(), |
|
]) |
|
|
|
|
|
class LoadImagesAndLabels(torch.utils.data.Dataset): |
|
|
|
def __init__(self, df, transforms, mode='train'): |
|
self.df = df |
|
self.transforms = transforms |
|
self.mode = mode |
|
|
|
def __len__(self): return len(self.df) |
|
|
|
def __getitem__(self,index): |
|
row = self.df.iloc[index] |
|
image_path = str(row.filename) |
|
image_path="/tmp/data/private_testset" |
|
image_path = os.path.join(images_root_path, str(row.filename)) |
|
img = Image.open(image_path).convert("RGB") |
|
img = np.array(img) |
|
|
|
if self.transforms is not None: |
|
img = self.transforms(image=img)['image'] |
|
|
|
if self.mode == 'test': |
|
return img |
|
|
|
label = torch.tensor(labels_class_map[row.binomial_name]).long() |
|
return img, label |
|
|
|
def get_corn_model(model_name, pretrained=True, **kwargs): |
|
model = timm.create_model(model_name, pretrained=pretrained, **kwargs) |
|
model = nn.Sequential( |
|
model, |
|
nn.Dropout(0.15), |
|
nn.Linear(model.num_classes, model.num_classes * 2) , |
|
nn.Linear(model.num_classes * 2, len(LABELS)) |
|
|
|
) |
|
return model |
|
|
|
class PytorchWorker: |
|
|
|
def __init__(self): |
|
self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") |
|
def _load_model(): |
|
print("Setting up Pytorch Model") |
|
|
|
print(f"Using devide: {self.device}") |
|
model = get_corn_model("vit_base_patch16_224", pretrained=False) |
|
model_ckpt = torch.load("./NB_EXP_V2_008/vit_base_patch16_224_224_bs32_ep16_lr6e05_wd0.05_mixup_cutmix_CV_0.pth", map_location=self.device) |
|
model.load_state_dict(model_ckpt) |
|
return model.to(self.device) |
|
|
|
self.model = _load_model() |
|
|
|
def predict_image(self, image: np.ndarray) -> list(): |
|
"""Run inference using ONNX runtime. |
|
:param image: Input image as numpy array. |
|
:return: A list with logits and confidences. |
|
""" |
|
image = image.to(self.device) |
|
outputs = self.model(image.unsqueeze(dim = 0)) |
|
logits = outputs |
|
return logits.tolist() |
|
|
|
|
|
def make_submission(test_metadata, model_path, model_name, output_csv_path="./submission.csv", images_root_path="/tmp/data/private_testset"): |
|
"""Make submission with given """ |
|
|
|
model = PytorchWorker() |
|
data = LoadImagesAndLabels(test_metadata, VALID_AUG, mode='test') |
|
predictions = [] |
|
for image in data: |
|
output = model.predict_image(image) |
|
string_label_dup = LABELS.get(str(np.argmax(output)), 'Acanthophis antarcticus') |
|
prediction_class = ORIGINAL_LABELS.get(string_label_dup, 1) |
|
predictions.append(np.argmax(prediction_class)) |
|
|
|
|
|
test_metadata["class_id"] = predictions |
|
|
|
user_pred_df = test_metadata.drop_duplicates("observation_id", keep="first") |
|
user_pred_df[["observation_id", "class_id"]].to_csv(output_csv_path, index=None) |
|
|
|
|
|
if __name__ == "__main__": |
|
|
|
import zipfile |
|
|
|
with zipfile.ZipFile("/tmp/data/private_testset.zip", 'r') as zip_ref: |
|
zip_ref.extractall("/tmp/data") |
|
|
|
MODEL_PATH = "pytorch_model.bin" |
|
MODEL_NAME = "swinv2_tiny_window16_256.ms_in1k" |
|
|
|
metadata_file_path = "./SnakeCLEF2024_TestMetadata.csv" |
|
test_metadata = pd.read_csv(metadata_file_path) |
|
|
|
make_submission( |
|
test_metadata=test_metadata, |
|
model_path=MODEL_PATH, |
|
model_name=MODEL_NAME |
|
) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
