image_video_upscaler

<h1>Go Image &amp; Video Upscaler</h1> <p>Project link: <a href="https://github.com/nalexand/video_upscaler">https://github.com/nalexand/video_upscaler</a></p> <p>This project is a high-performance, command-line tool written in Go for upscaling images and videos. It utilizes a custom, pre-trained model to perform a 3x upscale, with a focus on unblurring and denoising the source material. The processing is done in parallel to leverage multi-core CPUs for significant speed improvements.</p> <h2>Features</h2> <ul> <li><strong>3x Image Upscaling</strong>: Enhance the resolution of images (PNG) by a factor of 3.</li> <li><strong>3x Video Upscaling</strong>: Process videos by extracting frames, upscaling them individually, and reassembling them into a high-resolution video file.</li> <li><strong>Unblur &amp; Denoise</strong>: The included model is trained not just to enlarge, but also to correct common imperfections like blur and noise.</li> <li><strong>High Performance</strong>: Leverages Go's concurrency features to process image rows or video frames in parallel across all available CPU cores.</li> <li><strong>Command-Line Interface</strong>: Simple and clear flags for easy operation on local files.</li> <li><strong>Self-Contained Model</strong>: Uses a pre-trained model stored in a <code>.gob</code> file, which is Go's native binary format, download here: https://huggingface.co/datasets/nalexand/upscaler_unblur/tree/main</li> </ul> <h2>Prerequisites</h2> <p>Before using this tool, you must have the following software installed on your system:</p> <ol> <li> <strong>Go</strong>: The Go programming language (version 1.18 or newer recommended). You can download it from <a href="https://golang.org/" target="_blank" rel="noopener noreferrer">golang.org</a>. </li> <li> <strong>FFmpeg</strong>: A powerful multimedia framework used to extract frames from the input video and reassemble them with audio. <ul> <li>Installation instructions can be found on the <a href="https://ffmpeg.org/download.html" target="_blank" rel="noopener noreferrer">official FFmpeg website</a>.</li> <li>Ensure that both <code>ffmpeg</code> and <code>ffprobe</code> are available in your system's <code>PATH</code>. You can verify this by running <code>ffmpeg -version</code> and <code>ffprobe -version</code> in your terminal.</li> </ul> </li> </ol> <h2>Installation &amp; Setup</h2> <ol> <li> <strong>Clone the Repository</strong> <pre><code>git clone &lt;repository-url&gt; cd &lt;repository-directory&gt;</code></pre> </li> <li> <strong>Get the Model File</strong> <p>This program requires the model file <code>unblur_denoice_shared_model_x3_20.gob</code> to be present in the same directory as the executable. You can dowmload it here: https://huggingface.co/datasets/nalexand/upscaler_unblur/tree/main</p> </li> <li> <strong>Build the Executable</strong> <p>Create a binary from the source code. This command will produce an executable file (e.g., <code>unblur_server_binary</code> on Linux/macOS or <code>unblur_server_binary.exe</code> on Windows).</p> <pre><code>go build -o unblur_server_binary .</code></pre> </li> </ol> <h2>Usage</h2> <p>The tool is operated via command-line flags. The primary mode is <code>upscale</code>.</p> <h3>Image Upscaling</h3> <p>To upscale a single image:</p> <pre><code>./unblur_server_binary -mode=upscale -lr=input.png -out=output_upscaled.png</code></pre> <ul> <li><code>-mode=upscale</code>: Specifies the operation mode.</li> <li><code>-lr=&lt;path&gt;</code>: Path to the low-resolution input image.</li> <li><code>-out=&lt;path&gt;</code>: Path for the resulting high-resolution output image.</li> </ul> <h3>Video Upscaling</h3> <p>To upscale a video file:</p> <pre><code>./unblur_server_binary -mode=upscale -videoin=my_video.mp4 -videoout=my_video_upscaled.mp4</code></pre> <ul> <li><code>-videoin=&lt;path&gt;</code>: Path to the low-resolution input video.</li> <li><code>-videoout=&lt;path&gt;</code>: Path for the resulting high-resolution output video.</li> </ul> <p>The tool will automatically handle:</p> <ul> <li>Creating a temporary directory for frames.</li> <li>Extracting all frames from the input video using <code>ffmpeg</code>.</li> <li>Upscaling each frame in parallel.</li> <li>Reassembling the upscaled frames into a new video.</li> <li>Copying the audio track from the original video to the final output.</li> </ul> <h3>Using a Different Model</h3> <p>You can specify a different model file using the <code>-model</code> flag:</p> <pre><code>./unblur_server_binary -mode=upscale -lr=input.jpg -model=another_model.gob -out=output.png</code></pre> <hr>