Building traditional method for video recognition

Subject at PTIT: Hệ cơ sở dữ liệu đa phương tiện (Thầy Hóa)"

Method: SIFT_create

Pipline consists of feature extraction and finding similar videos

Extract Feature

Directory_path:

dataset

├── animals

    │ └── animals

        │ └── v1_animal.mp4

├── flowers

    │ └── flowers

        │ └── v1_flower.mp4

├── foods

    │ └── foods

        │ └── v1_food.mp4

├── humans

    │ └── humans

        │ └── v1_human.mp4

└── natures

    │ └─ natures

        │ └── v1_nature.mp4

read_files_in_directory(directory_path)

Search video

query_image is input image

Directory_feature:

features

├── animals

    │ └── v1_animal.mp4.npy

├── flowers

    │ └── v1_flower.mp4.npy

├── foods

    │ └── v1_food.mp4.npy

├── humans

    │ └── v1_human.mp4.npy

└── natures

    └── v1_fnature.mp4.npy

similar_videos_indices = find_similar_videos(query_image, directory_feature, k = 3)