@steevolicious so the exif tool is simply the way I used to view and update the data associated with each image. What I liked is that it could do a mass update on all the files quickly and easily.

I've got an Intel i7 10th gen with a RTX 2080, 32GB of RAM and an SSD. If you have a computer that has a dedicated graphics card and a half decent processor with some ram to back it up you should be able to run the program no problem. Might just take a bit of time.

unfortunately I don't own premier pro but I might use the free trial just to try this out!

Yes if you're drone is equipped with RTK GPS using GCPs are not strictly required

Hmm that's very odd. It could be that the SCP tool or QGis isn't cooperating with your machine. Any chance you're able to try it on a PC? I wasn't able to replicate the error. Maybe try saving the file in a different directory? Sorry I can't be more helpful.

Hi, unfortunately I don't own the data I used and I can't share it. On the Open Drone Map form there is a page that has all sorts of sample data sets opendronemap.org/odm/datasets/ The 10th one on the list named "helenenschacht" has both GCP data as well as RTK so you should be able to participate in the entire WebODM part of the tutorial but you wont be able to do the accuracy assessment. Hopefully this data is sufficient and best of luck!

Yes, the topo points are the elevations I collected with the rover GPS rather than the drone. I use them as my second set of data so that I can have a comparison with the drone data.

Thanks. I had the same confusion and question. Glad you cleared it up.

I haven't used the windows version but this thread could possibly have some answers for you. community.opendronemap.org/t/windows-docker-gpu/15209/4 It seems as though its possibly just broken and will not work or it could be a driver thing or possibly antivirus related. Best of luck hopefully you can get it working!

When you're using the GCP interface (at 12:14 in video) you use the GCP data you collected in the field. When you go to "select images and GCPs" (at 18:01 in video) to run ODM you use the GCP file that the GCP interface output. Hope this clears things up!

@@nathangeomatics thanks. I realize that you renamed the output file in the video. Thanks for the help.

Hi Jason, thanks for the great question. Unfortunately I never experimented with that but I'm pretty sure force GPS will disregard any GCPs. If you need to apply a bias correction I'd suggest bringing your data over to QGis and applying the correction there. In my second video on this subject I show how to do that at the 12min mark kzbin.info/www/bejne/pGS7Y5Kuj8aoZs0 I hope this helps!

kzbin.info/www/bejne/goWTlYx5nZ2jjtk This video should help! Other recourses are also found in the video's description.

WebODM download and install tutorial: github.com/OpenDroneMap/WebODM/ www.youtube.com/watch?v=LT0dY...

You're right, it is a pain; although not everyone has RTK capable drones and GCPs are the only option for some users.

I really like WebODM. It doesn't have all the features as metashape (yet) but for the price its hard to argue. I don't do this type of image processing every day, so its hard to justify the 179$ price tag for a personal license. With regard to litchi I've never used it but if it allows you to fly a regular grid and if you can trigger your camera at the right times I'm sure it will do just fine!

The RTK corrections need to be applied before bringing the data into ODM. I'll link a video that might be helpful. I included RTK in the title because for the purposes of a survey RTK GPS is required for good accuracy if GCP aren't being used. There isn't anything special you need to do in ODM to use RTK it will process your data either way. Sorry if this caused some confusion.

kzbin.info/www/bejne/nnizioeehrennMk

I used a DJI Phantom 4Pro modified by Klau Geomatics klauppk.com/UAV_Hardware/ Great question about the DOP. WebODM defaults to a DOP of 10m. This mean the maximum it will move an image in space is 10m, does not mean it will move the photo that much every time. This allows the program to move the photos as much as it wants to make them line up. Often times the amount it moves them is a whole lot closer to the actual DOP than the default. You can also assign the DOP inside the WebODM program which I did because for some reason it struggled to read in that exif data. With regard to your other questions, your probably better off using GCPs that have been measured accurately rather than rolling the dice with unknown accuracies and poor elevation data. Models created with GCPs can be pretty much as accurate as direct georeferencing provided you set up the GCPs well. Hopefully this answered your questions!

Unfortunately I dont own the dataset but I'll ask the owner and I'll see if I can get it to you.

​@@nathangeomatics I would like to process a similar dataset for my master thesis, so if you can ask him I'll appreciate.

In the example I used direct georeferencing to assign all x,y,z values. I simply showed the steps on how to use GCPs but I never actually ended up using them for georeferencing. The GCPs used in this portion of the tutorial were used to check for a bias and not georeferencing and thus weren't technically GCPs at all. I continued to call them GCPs simply because it allowed me to spilt my data into two data sets nicely without becoming too confusing. I needed one set of data for the bias correction and the other for the accuracy assessment. If you used GCPs for georeferencing you wouldn't need a bias correction because the values of the GCPs were used to assign the values to the imagery.

@@nathangeomatics thanks for the clarification. On a side note I've come to the belief that putting out dedicated targets in the field is of limited wisdom as it's often possible to use natural landmarks (manmade painted lines for parking or traffic control or cast concrete features for surveying), albeit it's probably a bit unorthodox. Which means that I can collect RTK GPS data after the fact and back-fit it into the ortho's. I guess the point I can glean from your tutorial is that I can use WebODM for generating a georeferenced DEM with low quality and conventional GPS data, and then I can do my adjustments with the RTK datapoints in QGIS instead of feeding them into WebODM. Does that make sense? (For a point of reference, I am new to this as a mechanical engineer and not a surveyor by education/trade/credential).

@@nathangeomatics can you also tell me about GCP limitations on 3D models generated in WebODM? I've been flying sites low and higher gimbal angle (say 125 feet and -55 below horizontal, plus at -80 for the ortho) hoping to improve upon my point clouds and 3d textured presentation. That being said, does it make sense to feed RTK GCP's in to help correct the overall geometric integrity, or does mapping the GCP points in to WebODM pictures taken at odd angles result in iffy results? Or does this really help a lot with elevation correction?

@@Project314 It is for sure possible to use existing features as GCPs and they would be just as valid but that could lead to confusion. For instance, using paint lines in a parking lot; its possible you might measure the wrong line intersection. Additionally, clear features aren't necessarily always available, for instance the survey could be in a pasture or field where there aren't any easily discernable features. Using a target will ensure you will always be able to get the exact center and it will be high contrast. You don't need targets but it will ensure consistency in your GCPs and will likely make the processing easier. You're also supposed to have an even distribution of GCPs across the whole study area and placing your own GCPs will guarentee that rather just relying on chance that you'll have enough evenly distributed distinct features. In short, you don't need to place targets but you probably should.

@@Project314 You don't need GCPs at all if your drone is equipped with a high precision GPS. I used them to help with an accuracy assessment. I didn't try two different angles being processed at the same time in WebODM, in order to get better coverage I flew a grid pattern so that photos were taken perpendicular to each other but the angle remained fixed. according to Meinen et al. (2020) a combination of nadir imagery and oblique 15-degree imagery is best but will have very similar accuracy to nadir imagery with a dense distribution of GCPs. James et al. (2020) however suggests 10 degree inclination because 15 degree may result in “promoting a previously unconsidered correlation between decentring camera lens distortion parameters and the radial terms known to be responsible for systematic topographic error”. I'm not totally sure how WebODM will handle two different angles and GCPs but you should try it. As long as your target is fully visible the angle shouldn't be a problem. If you're going through the effort to use GCPs anyway you could always head into QGis after the fact and preform a bias correction if deemed necessary. Meinen, B. U., & Robinson, D. T. (2020). Mapping erosion and deposition in an agricultural landscape: Optimization of UAV image acquisition schemes for SfM-MVS. Remote Sensing of Environment, 239, 111666. doi.org/10.1016/j.rse.2020.111666 James, M. R., Antoniazza, G., Robson, S., & Lane, S. N. (2020). Mitigating systematic error in topographic models for geomorphic change detection: Accuracy, precision and considerations beyond off‐nadir imagery. Earth Surface Processes and Landforms, 45(10), 2251-2271. doi.org/10.1002/esp.4878

I used an RTK GPS unit in conjunction with a base station. You can use pretty much whatever you want to assign x,y and z values to your GCPs such as a total station or a network connected RTK GPS (Cansel for example), the technology used will have an effect on the accuracy of the entire survey.

@@nathangeomatics Can I use a handheld GPS device to do the same. I am trying cut cost.

@@danpeters3903 You could try to use a handheld device but practically speaking that has an accuracy of +-3m. That means that your survey will have at the very best 3m accuracy. I would not recommend using a hand held device. If you absolutely need to use a handheld device take multiple measurements at each GCP and average those measurements in order to hopefully get something closer to the true value. For GCPs with the GPS I used it automatically took 180 measurements and averaged the shots. Doing 180 shots manually is of course not practical and probably not even useful with a handheld device so just take as many as you have time to do, I wouldn't go crazy maybe like 10 measurements in order to get a little closer. You should also do your best to place the phase center of the GPS antenna directly over the center of the GCP might not make a huge difference but might as well try. All that being said you likely won't get a very accurate survey but if all you need is +- 3m you should go for it! Let me know how it goes!

@@nathangeomatics many thanks for the tutorials, they're really helpful. Would appreciate your views on whether the on-board GPS on a mid-range DJI would be able to capture <3m accuracy if the drone was stationary on the ground for 5-10 minutes? I'm guessing that when the drone is moving at 30km/h the GPS reading and real-world camera location could understandably include error. If the on-board GPS can be quite accurate when static, would this make for a cheap and semi-accurate GCP specifier? Thanks.

​@@johnrogers3315 a stationary shot for 10min will yield much better results than capturing locations while its flying. Without RTK it will have limited accuracy but you should always do your best to make the most out of what you have available. Without RTK your absolute accuracy might not be great but the relative accuracy still could end up being quite good. Depending on your region there may be published data available for vertical control monuments and you may be able to place the DJI over a few different ones to get measurements allowing you to make a vertical correction to your vertical data if deemed necessary. I haven't tried using a hand held GPS or the onboard GPS on a DJI to do a survey but that is for sure something I want to try out next time I have a drone available. If you try it out let me know how it goes I'm quite interested to see the results.

Carleton University in Ottawa Ontario Canada!

Hopefully my profs who graded this thought the same lol

In this survey I had almost 300 images and when I didn't resize the program timed out and refused to run. I used ultra just because I could, no real good reason. I'm pretty new to WebODM and I just wanted to make a basic tutorial that could help someone walk through the entire process. I'm not an expert on the mechanics. Sorry I know that's not a very helpful answer but the ODM community might have better answers. I found these two discussion posts that might answer your questions about resizing, pc-quality and feature quality. community.opendronemap.org/t/what-are-the-tradeoffs-with-re-sizing-images/4813 community.opendronemap.org/t/question-regarding-feature-quality-and-pc-quality/6038 Have fun with WebODM! :)

Thanks for your questions, 1. For the GCPs I used EPSG:4326 which is in decimal degrees (Lat, Long) and meters for the elevation. The output however was in UTM coordinates so meters for x,y and z. That being said I'm pretty sure you can use any coordinate system provided you define it at the top of the text document as shown at 12:27 2. Unfortunately I cannot help you with this question, when I installed the program in September 2021 I just followed the directions provided at github.com/OpenDroneMap/WebODM/ and I didn't run into any complications. I remember the website saying that GPU acceleration wasn't quite ready yet so I didn't bother looking any further into it. Sorry. Hopefully this helped a little bit, enjoy WebODM!

@@nathangeomatics thanks, I'm new with ODM, I think it is interesting if somebody compare ODM with another tools (Pix4D, Agisoft, Drone Deploy, etc) for vertical accuracy with same dataset.