yolo_test.py

from ultralytics import YOLO
import cv2
import pandas as pd
import os
model=YOLO('runs/detect/train2/weights/best.pt')
results=model("bus.jpg", show=True)

barilist=[]
for r in results:
    boxes = r.boxes
    for box in boxes:
        x1, y1, x2, y2 = box.xyxy[0]
        # print(x1, y1, x2, y2)
        x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
        print(x1, y1, x2, y2)
        l = [x1, y1, x2, y2]
        barilist.append(l)
df = pd.DataFrame.from_records(barilist,columns=['scaled_xmin', 'scaled_ymin', 'scaled_xmax', 'scaled_ymax'])
df.to_csv("rafay.csv",index=False)


from keras.models import model_from_json
from sklearn.preprocessing import StandardScaler


model = "github/model@1535470106.json"
weights = "github/model@1535470106.h5"
csvfile_path = "rafay.csv"
results_dir = "github"



# get data
df_test = pd.read_csv(csvfile_path)
x_test = df_test[['scaled_xmin', 'scaled_ymin', 'scaled_xmax', 'scaled_ymax']].values

# standardized data
scalar = StandardScaler()
x_test = scalar.fit_transform(x_test)
scalar.fit_transform((df_test[['scaled_ymax']].values - df_test[['scaled_ymin']]) / 3)

# load json and create model
json_file = open(model, 'r')
loaded_model_json = json_file.read()
json_file.close()
loaded_model = model_from_json(loaded_model_json)

# load weights into new model
loaded_model.load_weights(weights)
print("Loaded model from disk")

# evaluate loaded model on test data
loaded_model.compile(loss='mean_squared_error', optimizer='adam')
distance_pred = loaded_model.predict(x_test)

# scale up predictions to original values
distance_pred = scalar.inverse_transform(distance_pred)

# save predictions
df_result = df_test
df_result['distance'] = -100000

for idx, row in df_result.iterrows():
    df_result.at[idx, 'distance'] = distance_pred[idx]
df_result.to_csv(os.path.join(results_dir, 'predictions.csv'), index=False)

cv2.waitKey(0)