1. If object has less texture: Keep a textured surface like magazine, book or newspaper under it. This will add enough features to images that they will get calibrated.
2. It’s best to capture images in an orientation, that is moving clockwise or anticlockwise around the object: This advice is based on test team experience with the product, as they found that randomly captured images always had subject overlap issue in neighbouring images.
3. Megapixels are hyped and not always the criterion for good reconstruction: High end SLR camera’s with wider lens only can capture enough optical information (light rays) that can be stored in 10 megapixel of space , lens of mobile devices or point and shoot cameras cannot collect that much of light rays to be stored in 10 megapixels of space. They mostly digitally interpolate , or add noise to convert low megapixel of image to a high megapixel image. So it’s best to keep your camera at low resolution (for mobile, point shoot cameras) as they are more accurate and won’t waste your precious computing time.
4. Number of images to be captured for best reconstruction : Mostly around 30 images are enough to create a good quality 3D model. Increasing number of images don't always guarantee a better model.
5. Make sure subject is static while images are captured: While you are moving around the object and capturing images, it’s very important that object stays still with respect to scene.
6. Well lighted subject: Outdoor scenes and subjects are usually well lighted by sun and the ambient light falling on them, but make sure the indoor subjects are well lighted.