Automatic image generator for Source Engine gamemaps (TF2)

This repository contains the tools used to generate a set of spectator images and level overview for every gamemap within TF2. It consists of two stages; the retrieval of the neccesary information from the .bsp files, and running the game to take the actual screenshots. The current dev gameshots system within TF2 is currently not sufficient for almost any community made gamemap.

The process to get from an .bsp file to an directory of screenshots of every gamemap consists out of two steps. The first step is extracting the required metadata from each bsp file and the second step is to run the game with that metadata.

This repository contains that was tailored for one specific setup (Windows storage + Linux with Steam via primusrun). It will need some small edits (e.g. installed additional packages via PIP, editing paths and maybe some small errors before it works). This is not an out-of-the-box solution.

Step 1) Metadata extractor for Source Engine Gamemaps (.bsp)

This metadata extractor accepts an folder with .bsp files and outputs a .json file containing all the extracted metadata for the maps that were processed.

Tested almost all TF2 maps available for download on the internet. It is combined with the minimal stubs VMF_Parser and MapStatsCoordinator from teamwork.tf to make this work.

Requirements/Dependencies

This metadata extractor relies on an external dependency called BSPSource. You need to make sure this is present on the same machine to make the data extraction work:

• Windows machine
• Python 2.7
• Java (JDK 1.6 or higher)
• BSPSource downloaded and placed inside the same folder.
• VMF_Parser in Python (included in the repository).

Example of the output

The output is a dictionary with every map name as an entry, with the metadata as the result.

"cp_glassworks_rc6a": {
"ammoPacks": [{"AutoMaterialize": "1", "StartDisabled": "0", "TeamNum": "0", "angles": "0 0 0", "classname": "item_ammopack_small", "fademindist": "-1", "id": "7817", "origin": "-3976 800 17"} ... ], 
"briefCases": [], 
"cameras": [{"StartDisabled": "0", "TeamNum": "0", "angles": "11 144 0", "associated_team_entity": "cp_red2", "classname": "info_observer_point", "defaultwelcome": "0", "fov": "0", "id": "753422", "origin": "-1600 1152 324"} ...], 
"containsHalloweenProps": false, 
"controlPoints": [{"classname": "team_control_point_master", "custom_position_x": "-1", "custom_position_y": "-1", "id": "6775", "origin": "160 -16 128.004", "targetname": "5cp_master", "team_base_icon_2": "sprites/obj_icons/icon_base_red", "team_base_icon_3": "sprites/obj_icons/icon_base_blu"} ...], 
"dimensions": [[-5466.0, -4475.0, -744.0], [5466.0, 4338.0, 1428.0], {"scale": 9, "screenHeight": 1080, "screenWidth": 1920}, {"x": 0.0, "y": -68.5, "z": 1428.0}], 
"entityCount": 3756, 
"fileHash": "9f169ea81bc9a1b2c6d4f5f0bdbde7d0", 
"healthKits": [{"AutoMaterialize": "1", "StartDisabled": "0", "TeamNum": "0", "angles": "0 0 0", "classname": "item_healthkit_small", "fademindist": "-1", "id": "7821", "origin": "-4024 800 17"} ...], 
"normalizedMapName": "cp_glassworks", 
"resupplyLockers": [5270.41, -897.867, 8.16007], 
"skyboxCamera": {"angles": "0 0 0", "classname": "sky_camera", "fogblend": "0", "fogcolor": "113 115 142", "fogcolor2": "255 255 255", "fogdir": "1 0 0", "fogenable": "1", "fogend": "8000", "fogmaxdensity": ".9", "fogstart": "100", "id": "1016124", "origin": "88 9672 427.5", "scale": "16", "use_angles": "0"}, 
"spawnPoints": [{"StartDisabled": "0", "TeamNum": "2", "angles": "0 0 0", "classname": "info_player_teamspawn", "controlpoint": "cp_red1", "id": "5459", "origin": "-5328 1120 24"} ...], 
"tournamentStage": null, 
"version": 3869
}

Getting started creating metadata

• Clone this repository.
• Download and install Java, and locate the java.exe binary.
• Download and install Python 2 or 3.
• Download BSPSource and place the .jar in the same folder as you just checked out from this repository.
• Open up map_data_gatherer.py in your favorite text Editor and change the JAVA_EXEC and MAPS_DIR according to your system. The MAPS_DIR is the location where the .bsp files are located you want to extract metadata from.
• Open command line, navigate to this repository on your local disk and execute python map_data_gatherer.py.
• Wait for completion, there should appear a json file in your ./metadata folder.

Step 2) Running the game to gather screenshots

Armed with an folder full of gamemaps, and an .json file containing all the relevant metadata about that map, you can now generate screenshots for TF2. This is designed for Linux only, as linux allows finer control over subprocesses run on the system. It might work on windows, but you would need to edit the code significantly.

Requirements/Dependencies

• Linux machine (tested for Debian/Ubuntu)
• Nvidia GPU (and primusrun, can be removed with some small code edits)
• Python 2.7
• Packages: pyautogui, numpy, PIL, psutil
• Team Fortress 2 installed on your machine, via Steam.
• 1920x1080 screen (can be edited in the code: search for 1920 in the map_data_gatherer).

Example of the output

For each gamemap that it processes it creates the following results:

• An screenshot of each spectator camera in the gamemap (spectator_{camera}.png).
• An overview image of the gamemap, based on the estimated center of the map, merged from 5 different screenshots (overview_merged.png).
• An metadata.json file containing all the information about the map we just processed.

Getting started creating screenshots

• Make sure console is enabled in Team Fortress 2, and it can launch properly via Steam (no pending updates etc.).
• Place the gamemaps you want to make screenshots of, in the games directory under tf\maps\.
• Edit run_game_map.py and set the correct paths on where to find the metadata file you created in step 1 and where to store the results.
• Make sure you have Steam open and logged in, but did not start TF2 yet.
• Open a terminal and run python run_game_map.py.
• Wait for the magic to happen, and wait untill it closes the game.
• The screenshots are now in the given folder.