English version
Objective
This guide documents an end-to-end workflow for preparing camera trap data:
rename_images() -> process_metadata() -> generate_plot() (optional) -> process_training_images() -> copy_in_batches().
Inputs, outputs, and generated files by stage
| Stage | Minimum input | In-memory output | Files written |
|---|---|---|---|
rename_images() |
input_dir containing images |
Result tibble (or a list with plan +
result) |
Creates a site folder and copies renamed files |
process_metadata() |
Folder containing renamed images | Metadata tibble | Writes a metadata CSV |
generate_plot() |
Data frame and columns used for aesthetics | ggplot object |
None |
process_training_images() |
CSV with Directory and FileName |
Sample tibble | Creates dest_dir, writes sample and annotation CSVs,
and optionally copies images |
copy_in_batches() |
Data frame with id and path |
Report data frame | Creates batch folders and copies files (or simulates the operation
with dry_run) |
Reproducible pipeline with temporary paths
library(cam2model)
# Temporary paths
raw_dir <- tempfile("cam2model_raw_")
renamed_dir <- tempfile("cam2model_renamed_")
train_dir <- tempfile("cam2model_train_")
batch_dir <- tempfile("cam2model_batches_")
dir.create(raw_dir, recursive = TRUE)
dir.create(renamed_dir, recursive = TRUE)
dir.create(train_dir, recursive = TRUE)
# NOTE: rename_images() and process_metadata() require real images (EXIF).
# These calls are shown for reference; the following stages use synthetic data.
# renamed <- rename_images(
# input_dir = raw_dir,
# output_dir = renamed_dir,
# site_name = "demo_site"
# )
# metadata <- process_metadata(
# output_dir = renamed_dir,
# output_file = "metadata.csv",
# UserComment = FALSE
# )
# ------------------------------------------------------------------
# Minimal synthetic metadata to demonstrate process_training_images()
# ------------------------------------------------------------------
src_dir <- tempfile("cam2model_src_")
dir.create(src_dir, recursive = TRUE)
img1 <- file.path(src_dir, "img1.jpg")
img2 <- file.path(src_dir, "img2.jpg")
writeBin(charToRaw("x"), img1)
writeBin(charToRaw("y"), img2)
metadata_tbl <- data.frame(
Directory = c(src_dir, src_dir),
FileName = c("img1.jpg", "img2.jpg"),
File_hour = c(9, 14),
stringsAsFactors = FALSE
)
metadata_csv <- tempfile("cam2model_metadata_", fileext = ".csv")
readr::write_csv(metadata_tbl, metadata_csv)
sampled <- process_training_images(
metadata_file = metadata_csv,
dest_dir = train_dir,
output_sample_file = "sampled_images.csv",
output_annotation_file = "annotation_data.csv",
sample_proportion = 1,
copy_images = FALSE,
show_progress = FALSE
)
#> Rows: 2 Columns: 3
#> ── Column specification ────────────────────────────────────────────────────────
#> Delimiter: ","
#> chr (2): Directory, FileName
#> dbl (1): File_hour
#>
#> ℹ Use `spec()` to retrieve the full column specification for this data.
#> ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
sampled
#> # A tibble: 2 × 4
#> Directory FileName File_hour file_path
#> <chr> <chr> <dbl> <chr>
#> 1 /tmp/RtmpU5bHnX/cam2model_src_6e3457b0d066 img1.jpg 9 /tmp/RtmpU5bHnX…
#> 2 /tmp/RtmpU5bHnX/cam2model_src_6e3457b0d066 img2.jpg 14 /tmp/RtmpU5bHnX…Optional plot
plot_data <- data.frame(
Image_dttm = as.POSIXct("2024-08-01 00:00:00", tz = "UTC") + 0:3 * 3600,
bLuma = c(120, 130, 125, 140),
Camera = c("C1", "C1", "C2", "C2")
)
generate_plot(
processed_metadata = plot_data,
x = Image_dttm,
y = bLuma,
color = Camera,
facets = Camera
)
Batch copying (dry run)
batch_report <- copy_in_batches(
df = data.frame(
id = sprintf("img_%03d", seq_len(nrow(sampled))),
path = sampled$file_path,
stringsAsFactors = FALSE
),
output_dir = batch_dir,
batch_size = 250L,
dry_run = TRUE,
verbose = FALSE
)
batch_report[, c("id", "batch_folder_name", "dest_file_name", "status")]
#> id batch_folder_name dest_file_name status
#> 1 img_001 batch_001 img_001_img1.jpg dry_run
#> 2 img_002 batch_001 img_002_img2.jpg dry_run