Photogrammetry is the art of reconstructing digital 3D models using photos.
To reconstruct a 3D model of an object, photos are captured from different angles.

Tips for capturing photos

A set of images can be extracted from a video too. Note that
a set of photos has advantages over a set of extracted images from a video:
Photos contain meta information like focal length and sensor width.
Photos tend to be less blurry. 

ffmpeg -i "$VIDEO" -r "$FPS" -qmin 1 -qscale:v 1 img_%04d.jpg

COLMAP

Put all photos into a folder called images inside a named project folder: 

PROJECT="my_photogrammetry_experiment"
mkdir -p "$PROJECT/images"
cp *.jpg "$PROJECT/images"

Optionally (but highly recommended), create image masks for each image.
Image masks tell colmap where to look for keypoints during feature extraction.
For an image images/0123.jpg, the corresponding mask has to be named masks/0123.jpg.png.
No features will be extracted in regions, where the mask is black (pixel value 0).

Execute the reconstruction: 

colmap automatic_reconstructor \
  --workspace_path "$PROJECT" \
  --image_path "$PROJECT/images" \
  --mask_path "$PROJECT/masks"

The reconstructed point cloud is written to: 

$PROJECT/dense/0/fused.ply

The reconstructed mesh is written to: 

$PROJECT/dense/0/meshed-poisson.ply

MeshLab

MeshLab can import point clouds and meshes stored in the Stanford Polygon File Format (*.ply): 

File > Import Mesh...

Using the point cloud, a mesh can be computed using the following filter: 

Filters > Remeshing, Simplification and Reconstruction > Surface Reconstruction: Screened Poisson
Reconstruction Depth: 12

There is also a filter for mesh simplification.
The mesh should be smoothed before simplification. 

Filters > Remeshing, Simplification and Reconstruction > Simplification: Quadric Edge Collapse Decimation
Target number of faces

Blender

File > Import > Stanford PLY (.ply)

Sculpting
In the toolbar at the bottom: Smooth

External Links