If you ask a random group of photographers what their favorite things about photography are, odds are “spending hours keywording images” is not on anyone’s list! Keywording is a task that is easy to put off. I am guilty of this and quite often behind getting it done. I’ve set aside a specific “keywording” day on the calendar, used keyword lists to make it easier, I’ve even adked my wife to hide my memory cards until I get the latest batch of images done. For all of this, keywording remains my number one most procrastinated task on my photographic to-do list.
During my eternal struggle to battle and overcome the “keywording monster”, I recently found a plugin in the Adobe library that caught my eye. “Cloud Tagger“, uses Google’s “Cloud Vision API” platform to search for, find and suggest keywords for your photos based on their visual characteristics. It analyzes your photo then matches it with keywords pulled from similar images across the web.
There have been a number of quiet mutterings over the past few years that we may see this type of search feature in Lightroom one day. It might be part of the much talked about “Project Nimbus.” Cloud Tagger makes it a reality now. It is not a free plugin, but for $20.00, I was ready to test it in battle against my ever present and looming “keywording monster”!
Installation and Use
Cloud Tagger is available in the Adobe Add-ons library. That makes installation simple. Purchase the plugin. Adobe Creative Cloud automatically installs it in your version of Lightroom . Restart Lightroom. Navigate to File > Plug-in Manager, then activate Cloud Tagger.
To use the plugin, select an image or group of images and go to Lightroom’s Library menu > Plug-in Extras > Get keyword suggestions.
NOTE: You must be in the Library module and have at least one image selected for Cloud Tagger to work!
Once the plugin runs, a window will pop up showing any current keywords you have entered along with suggestions from Google with a percentage number, called the “confidence factor.” The confidence factor is Google’s estimated accuracy of the keyword match. The plugin will auto-select any keywords above a user determined threshold. Cloud Tagger defaults to 85%. I tried a few different thresholds, finally settling for 60%. Change the threshold under File > Plug-in Manager > edit Cloud Tagger’s settings.
So How Good Is It?
I ran Cloud Tagger through a variety of images, from barred owls to my son running cross-country. Overall, I was happy with the results. Of course there are a few exceptions. Below are some examples to show how well it performed with different subjects, scenery and conditions.
Round 1 – Animals
Keywording animals can be tricky, many animals are difficult to identify in terms of species, especially when it comes to birds, insects, etc. I wasn’t expecting Google to give consistent species ID’s, and generally that was true. There were also some incorrect ID’s, usually due to color or body shape similarities. For wildlife photos, while it did not consistently offer species specific keywords, it usually got the general type of animal correct and suggested many useful terms.
While it handled the wild horse mare and foal image well, the gray fox photo was not identified by species. “Chipmunk” actually had a higher confidence factor than fox!
Round 2 – Birds
Birds are tough
because they are often identified by subtle characteristics. So I won’t be too critical of Cloud Tagger’s performance here. I ran several species through it. Often, it was right on the bird type (birds of prey, shorebirds, etc.) it was rarely able to ID the species (suggesting just “owl” instead of “barred owl”). Three different photos of a barred owl produced many relevant keywords. It got a few things wrong. A barred owl classified as Accipitriformes instead of Strigiformes?! My inner birder wept a little!
Round 3 – Landscapes
I was very happy with this category, Cloud Tagger generally nailed it in terms of providing accurate keywords for all types of landscape images. Most of the time it got the type of landscape right (mountain, beach, etc.), as well as time of day (sunrise, dusk, etc.).
Both images received solid relevant results, including landforms, weather, time of day, and more. Keywording high-fives all around!
Round 4 – Vehicles
I tried this one just for fun. I was curious to see how well Cloud Tagger would do. For the old Ford photo in black and white, it did impressively well. It not only identifying it was a vehicle, it also pegged it as an antique one. This was a result I found on most images that included the whole vehicle. For the pink van, I was curious if it would recognize the classic “VW” hood ornament. Unfortunately, it did not. However, I suspect this was due mostly to the low contrast between the ornament and the VW bus’ body. Or, as I mentioned earlier, maybe due to its difficulty in distinguishing different types of “bugs” (Ba-dum-tss).
Round 5 – People
Cloud Tagger also performed extremely well in this category, providing detailed and relevant keywords for most images I experimented with. In the western image it identified not just “cowboy” and “horse” but also that one was a “pack animal”. Likewise, it would be nearly impossible for it to know that my son was running a cross-country race instead of a sprint. It did identify the activity of running, that it was a sport, and that he was wearing a yellow team jersey. Pretty impressive!
Round 6 – Low Light
I had a feeling this would be a mixed bag. A known weakness of Google’s Cloud Vision is for dark or low contrast images. Here I saw the biggest fail and “head scratcher”, identifying cowboys around a campfire as a vehicle. I was pleasantly surprised the marina shot produced a plethora of keywords involving boats, time of day (“night”, “evening”) and place (“harbor”, “marina”).
The Good Stuff
The plugin is stable, fairly fast, and well supported. Shortly after purchasing “Cloud Tagger”, I emailed a few questions to the plugin’s creator on its use. I was not expecting a fast response, most of these plugins are not produced by big companies with loads of staff, but by people in their spare time. Nonetheless, I received a quick reply, with a promise to follow-up. It’s good to see this project is being supported and improved by its creator.
From a time-saving standpoint, my official guesstimate is it saved about half the time I would normally spend keywording. I picked 20 shots at random, keywording them as I normally would. Looking up species, relying on existing keywords, doing searches for subject details, etc. I then picked 20 more random images and started keywording each using Cloud Tagger. Often, I still had to manually add a few keywords. However, the time-savings and extra keyword suggestions that may not have occurred to me, made the plugin very useful.
The Wish List
Many Lightroom plugins are “interpreters”. They allow Lightroom to communicate and interact with other software or services. That is exactly what Cloud Tagger does. The few issues I had using the plugin are actually items for Google and Adobe to address, not necessarily the plugin author.
As I expected, the biggest struggle Google’s Cloud Vision had was with dark images. For wildlife, it was rarely successful at identifying a specific species, but could usually identify the general family (bear, butterfly, etc.). Since this is a project under development, I would expect it to get better over time. Google is, in my opinion, one of the smarter and more innovative companies in business, so I would expect this project to only improve.
On the Adobe side, a feature I would like for PC users, such as myself, is the ability to add items to the Library right-click context menu, and/or create your own custom keyboard shortcuts. This is possible on Macs, but PC folks have to use a third-party plugin or script. To access Cloud Tagger you have to go through the “Library” drop down menu, which slows down the workflow. Sure, it’s only a few seconds, but that adds up when doing large batches of images. I am a big fan of keyboard shortcuts. The ability to map the plugin to a key combination would be an excellent time saver. This is an often requested feature for PC users. Hopefully, it will appear in a future version of Lightroom.
The Bottom Line
Any time you buy something for your photography, you should know exactly how it’s going to help you, and that it is worth the cost for the benefits it provides. In my opinion, spending twenty bucks for saving untold hours at a task I normally enjoy about as much as changing my cats’ litter box is a bargain .
The Cloud Tagger plugin helped me add large numbers of relevant keywords in a short time period. The only area I really found it lacking in was “specificity”. Simply put, Google’s Cloud Vision doesn’t know the story behind the photo. Without this it will struggle to be specific on locations, species, or some activities. While I do believe this will improve greatly in the future, it can’t know all the details of your shoot by reading your mind…yet. For now, you will still have to add some of these specific keywords yourself.
“Cloud Tagger” is something I am already putting to good use in my workflow, and I am happy I found it. I hope it helps all of you defeat your “keywording monster” as well!
When not writing about himself in the third person, he enjoys sunsets and long walks on the beach while carrying 40 pounds of camera gear. He can most often be found wading through a swamp, hunting down a good burger joint, or enjoying time with in the great outdoors.
Latest posts by Jason Hahn (see all)
- How to make beautiful backgrounds with v-flats - May 18, 2019
- How to use negative fill to create deep, dramatic photos - April 3, 2019
- Keeping your photography equipment safe around sand and salt water - March 21, 2019