This is the place where you should post your entries to the 3D model of industrial site competition.
The model or video must be posted as a comment to this post in order to be considered as the competition entry. However, feel free to share it on your social media or in our Capturing Reality Arena.
Feel free to comment, admire, give advice, but please be polite and stick to the rules of the forum and the competition****.
We are looking forward to seeing many amazing videos and models.
For those who have just stumbled across this group:
The aim of this competition is to create a geo-referenced 3D model of an industrial site.
RealityCapture software must be used.
No post-processing with other tools.
The 3D model must be geo-referenced.
Here you can find more information on the competition and how to participate.
Images collected using DJI Inspire 1 Pro drone, 247 x 16mp images, Birkdale in UK, site of old building, data inc GCPs collected for a survey company for input to their CAD software in Nov 2016. Data set now put through Reality Capture trial here, took 35 mins to calculate the model. https://youtu.be/orpk3i8lX8I
This is my entry for RealityCaptures’ Industrial Site Competition.
Been UAV surveying/mapping for a couple of years testing various software but not until recently I got to ‘‘play’’ with RealityCapture!
So here is part of the final product of a small research/project conducted with a geologist colleague of mine;
We used a DJI P4P to capture and survey a mines’ entire bench (1.5km long) face within one batterys’ time while the UAV was moving sideways and constantly at a certain height above ground using known existing DEM.
Our goal was to extract every different geological layer (aka strata) from the 3D product and combine the data acquired with the drillhole data in order to enhance the geological model.
Data acquired only with a DJI Phantom 4 Pro
455 images (5472x3648)
Approximately 20 minutes of flight time
3D Model consisting of 364.7million triangles, 182 parts reconstructed in High Detail
Sketchfab Model, in really low resolution due to 50mb file size limitations even though I chopped up the original 3D model from 1.5km to 200m; https://skfb.ly/6MvOZ
Model of “Tampere Deck Arena” worksite of enormous building complex by big construction office “SRV” in Finland.
Worksite is at this moment only small portion of model but my scan is captured from complete project area.
1053 images were captured with a DJI Phantom 4 Pro - drone. Flight was automated with UgCS - photogrammetry tool by 85% x 85% overlapping dual grid flight mission from 160 meters.
Phantom 4 Pro camera takes 5464 x 3640 // 19.8 MP raw images which results as 38MB raw files. images were postprocessed in Adobe Lightroom resulting in 11.4 MB JPG:s –> therefore overall processed data were roughly 1,2GB.
Flight time:
1H. 26Min // 4 flights
Image post processing and exporting in Lightroom: 40 min.
RelityCapture processing info from sketchfab export: 3h, 12m and 12s
This is my application for the competition. I used DJI phantom, and a Nikon D3500. I used a combination of ground based photos, drone photos and and video captures.
** NB I updated my entry a little to suit the new deadline **
I captured about 1500 photos for this. the mesh creation took around 6 hours and texture creation and a similar amount of time, total time for creation was around 15 hours for the full scene which I then cut down. Due to the competition restrictions (namely Sketchfab) I only created a section of the capture and not the whole environment.
My original mesh was over 270,000,0000 triangles and has been heavily reduced for Sketchfab and this competition. This mesh was geolocated via the DJI drone GPS system.
My entry to the 3D industrial site competition is an old historic mine near Silverton, Colorado. The Upper Cement Creek Mine is an inactive mine and is part of a reclamation project. Processed in High Detail mode.
hello There. I decided to join to the competition and this object is my entry. This coal mine was scanned by me time ago. I just bought brand new Dji Phantom 4 Pro and I was looking for perfect object for scanning. Every day when I drove to work I passed by this mine. This reconstruction isn’t perfect. In that period I was getting new experience, trying different ways of scanning from the air. This scan is one of first my scans processed in Reality Capture. to create this scan I use pictures only from P4P. It was impossible take a pictures from the ground.
This object was created from 1766 pictures
processing time:
Here is my entry into the RC industrial competition.
This quarry wall was created for the purpose of safe and fast Geotechnical investigation and analysis. the model was created using a mixture of terrestrial lidar from ground level and drone photogrammetry from above. 1754 cameras were used in the making of this model and what has been uploaded is less than half. utilising the terrestrial lidar we are able to create a mesh with an accuracy of +/-5mm, creating a very precise mesh for the geotechnical team to carry out their analysis.
Having only just found out about this competition, My exports are a little rushed. Not sure why my video has come out so strange but, it took about 10 hour to export, still learning my way round RC at the mo.
Here is my Chevelon Canyon Bridge entry. The Chevelon Canyon Bridge is located in Navajo County, Arizona USA. It is a Pony truss bridge spanning of the over Chevelon Creek and was built-in 1913. It has a total length of 102.7 ft (30.3 M). The bridge is posted in the National Register of Historic Places. It is still in operation today but only sees an average of 36 vehicles/day.\
This bridge has been on my bucket list for bridges to scan. I first found out about it through the bridgehunter.com website and thought that it would make a great 3D model. I drove over 800 kilometers to get the images and only had 3 hours to do it. The model did not come out as well as planned due to the bridge being freshly painted with a shiny high gloss black paint and the 30+ Km/h winds. UGH! I plan to revisit in the future with more time to capture the bridge trusses better.
A DJI Phantom 4 Pro drone was used to capture the 1800+ 20MP images using a combination of free-flight mode, point of interest mode, and Pix 4D’s capture app for the image acquisition. The captured images were pre-processed using Photoshop Lightroom using the following adjustments:
Decrease the highlights -90
Increase the shadows +90
Decrease white level so that there is no clipping of the highlights
Increase black level so that there is no clipping of the shadows
Increase clarity to 25
Increase vibrancy to 20
Increase saturation to 5
Increase sharpening slightly
Masking increased to prevent sharpening of the highlighted areas
The images were exported and added to Reality Capture and processed in High Detail mode. Overall processing time 2d 06h:49m:51s
)
