icputrd/Inception-V3-Human-Bodypart-Classifier

1---2pipeline_tag: image-classification3license: apache-2.04---5 6# Model Card: Fine-Tuned InceptionV3 for Human Bodypart Image Classification7 8This CNN model was developed to perform human bodypart classification for forensic purposes.9 10## Model Details11 12### Model Description13 14- **Funded by:** National Institute of Justice15- **Model type:** CNNs for Image Classification16- **Base Model:** InceptionV3 pretrained on ImageNet17 18## Dataset19 20- Dataset Name: Human Decomposition Image Dataset21- Source: The dataset used in this study was obtained from the Forensic Anthropology Center (FAC) at the University of Tennessee, Knoxville, but due to privacy considerations, it is not available for public access. Please reach out to obtain access.22- Classes: arm, hand, foot, legs, fullbody, head, backside, torso, stake, and plastic. stake and plastic classes were23included for filtering out images where bodyparts are covered with plastic or images showing stake with unanonymized donor IDs,24which is often the case in forensic anthropology.25 26## Usage27```python28from tensorflow.keras.models import load_model29import numpy as np30from tensorflow.keras.preprocessing.image import img_to_array, load_img31 32# Load the entire model33model = load_model('inception_acc_0.989001-_val_acc_0.98252.h5')  34 35# Load and preprocess an image36img = load_img('path_to_image.jpg', target_size=(299, 299))  # adjust size as per model input37img = img_to_array(img)  # convert to numpy array38img = np.expand_dims(img, axis=0)  # add batch dimension39img = img / 255.0  # normalize pixel values if needed40 41# Make predictions42predictions = model.predict(img)43 44# Use argmax to get the class label 45predicted_class = np.argmax(predictions, axis=1)46```