im2txt on webcamera

Description

Show and Tell realtime detection with webcamera based on "run_inference.py".

Additional requirement

• OpenCV
  • to use cv2.VideoCapture()

Installation

1. Clone this repository.

2. Copy "im2txt" folder and "process_camera.sh" to "~/models/research/im2txt".

3. Open "~/models/research/im2txt/im2txt/BUILD" and add this comment.

py_binary(
    name = "run_inference_camera",
    srcs = ["run_inference_camera.py"],
    srcs_version = "PY2AND3",
    deps = [
        ":configuration",
        ":inference_wrapper",
        "//im2txt/inference_utils:caption_generator",
        "//im2txt/inference_utils:vocabulary",
    ],
)

4. Run this command at "~/models/research/im2txt".

bazel build -c opt //im2txt:run_inference_camera

How to use

1. Open "process_camera.sh" and edit line to your path.

--checkpoint_path="${HOME}/im2txt/model/train" \
--vocab_file="${HOME}/im2txt/data/mscoco/word_counts.txt" \

2. Run this command at "~/models/research/im2txt".

chmod +x process_camera.sh

3. Now you can run.

./process_camera.sh

Other

I can answer questions in English or Japanese.

日本語の質問でも大丈夫ですよ！